{
  "schema_version": "1",
  "generated_at": "2026-07-17T09:10:39.557213+00:00",
  "vllm_version": "0.25.1",
  "vllm_recipes_sha": "5d1bf23",
  "accelerator_brands": {
    "h100": "nvidia",
    "h200": "nvidia",
    "b200": "nvidia",
    "gb200": "nvidia",
    "b300": "nvidia",
    "gb300": "nvidia",
    "dgx_station_gb300": "nvidia",
    "dgx_spark_gb10": "nvidia",
    "mi300x": "amd",
    "mi325x": "amd",
    "mi355x": "amd",
    "trillium": "google",
    "ironwood": "google",
    "xeon6": "intel",
    "xeon5": "intel",
    "rtxpro-6000-blackwell-server-edition": "nvidia"
  },
  "index": {
    "h100": {
      "google/diffusiongemma-26B-A4B-it": [
        {
          "name": "DiffusionGemma 26B-A4B IT",
          "huggingface_id": "google/diffusiongemma-26B-A4B-it",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/diffusiongemma-26B-A4B-it-FP8-dynamic": [
        {
          "name": "DiffusionGemma 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/diffusiongemma-26B-A4B-it-FP8-dynamic",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/diffusiongemma-26B-A4B-it-NVFP4": [
        {
          "name": "DiffusionGemma 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/diffusiongemma-26B-A4B-it-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-12B-it": [
        {
          "name": "Gemma 4 12B IT",
          "huggingface_id": "google/gemma-4-12B-it",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-12B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "google/gemma-4-12B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 12B IT (w4a16)",
          "huggingface_id": "google/gemma-4-12B-it-qat-w4a16-ct",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-12B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-26B-A4B-it": [
        {
          "name": "Gemma 4 26B-A4B IT",
          "huggingface_id": "google/gemma-4-26B-A4B-it",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it": [
        {
          "name": "Gemma 4 31B IT",
          "huggingface_id": "google/gemma-4-31B-it",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-31B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 31B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-31B-it-FP8-dynamic",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/gemma-4-31B-it-NVFP4": [
        {
          "name": "Gemma 4 31B IT (nvfp4)",
          "huggingface_id": "nvidia/gemma-4-31B-it-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 31B IT (w4a16)",
          "huggingface_id": "google/gemma-4-31B-it-qat-w4a16-ct",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it": [
        {
          "name": "Gemma 4 E2B IT",
          "huggingface_id": "google/gemma-4-E2B-it",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E2B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E2B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E2B-it-FP8-dynamic",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E2B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E2B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-w4a16-ct",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E2B IT (mobile)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-mobile-ct",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "google/gemma-4-E4B-it": [
        {
          "name": "Gemma 4 E4B IT",
          "huggingface_id": "google/gemma-4-E4B-it",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E4B-it-FP8-dynamic",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E4B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E4B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-w4a16-ct",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E4B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E4B IT (mobile)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-mobile-ct",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Base": [
        {
          "name": "LFM2.5 1.2B Base",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Base",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Instruct": [
        {
          "name": "LFM2.5 1.2B Instruct",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Instruct",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-JP-202606": [
        {
          "name": "LFM2.5 1.2B JP (202606)",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-JP-202606",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-JP": [
        {
          "name": "LFM2.5 1.2B JP",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-JP",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Thinking": [
        {
          "name": "LFM2.5 1.2B Thinking",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Thinking",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-230M": [
        {
          "name": "LFM2.5 230M",
          "huggingface_id": "LiquidAI/LFM2.5-230M",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-350M": [
        {
          "name": "LFM2.5 350M",
          "huggingface_id": "LiquidAI/LFM2.5-350M",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-8B-A1B": [
        {
          "name": "LFM2.5 8B A1B",
          "huggingface_id": "LiquidAI/LFM2.5-8B-A1B",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-VL-1.6B": [
        {
          "name": "LFM2.5 VL 1.6B",
          "huggingface_id": "LiquidAI/LFM2.5-VL-1.6B",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-VL-450M": [
        {
          "name": "LFM2.5 VL 450M",
          "huggingface_id": "LiquidAI/LFM2.5-VL-450M",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M2.1": [
        {
          "name": "MiniMax-M2.1",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.1",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M2.5": [
        {
          "name": "MiniMax-M2.5",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.5",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--compilation-config": "{\"cudagraph_mode\":\"PIECEWISE\"}",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "32768",
                "--kv-cache-dtype": "fp8",
                "--moe-backend": "triton",
                "--attention-backend": "FLASHINFER",
                "--enable-flashinfer-autotune": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "PYTHONNOUSERSITE": "1",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0",
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--compilation-config": "{\"cudagraph_mode\":\"PIECEWISE\"}",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "32768",
                "--kv-cache-dtype": "fp8",
                "--moe-backend": "triton",
                "--attention-backend": "FLASHINFER",
                "--enable-flashinfer-autotune": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "PYTHONNOUSERSITE": "1",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0",
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.5-NVFP4": [
        {
          "name": "MiniMax-M2.5 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.5-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--compilation-config": "{\"cudagraph_mode\":\"PIECEWISE\"}",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "32768",
                "--kv-cache-dtype": "fp8",
                "--moe-backend": "triton",
                "--attention-backend": "FLASHINFER",
                "--enable-flashinfer-autotune": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "PYTHONNOUSERSITE": "1",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0",
                "VLLM_FLOAT32_MATMUL_PRECISION": "high",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--compilation-config": "{\"cudagraph_mode\":\"PIECEWISE\"}",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "32768",
                "--kv-cache-dtype": "fp8",
                "--moe-backend": "triton",
                "--attention-backend": "FLASHINFER",
                "--enable-flashinfer-autotune": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "PYTHONNOUSERSITE": "1",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0",
                "VLLM_FLOAT32_MATMUL_PRECISION": "high",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2.7": [
        {
          "name": "MiniMax-M2.7",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.7",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.7-NVFP4": [
        {
          "name": "MiniMax-M2.7 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.7-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2": [
        {
          "name": "MiniMax-M2",
          "huggingface_id": "MiniMaxAI/MiniMax-M2",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/MiniMax-M2-NVFP4": [
        {
          "name": "MiniMax-M2 (nvfp4)",
          "huggingface_id": "RedHatAI/MiniMax-M2-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "OpenGVLab/InternVL3_5-8B": [
        {
          "name": "InternVL3.5",
          "huggingface_id": "OpenGVLab/InternVL3_5-8B",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B": [
        {
          "name": "Qwen3.6-35B-A3B",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B-FP8": [
        {
          "name": "Qwen3.6-35B-A3B (fp8)",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B-FP8",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3.6-35B-A3B-NVFP4": [
        {
          "name": "Qwen3.6-35B-A3B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.6-35B-A3B-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3.2": [
        {
          "name": "DeepSeek-V3.2",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.2",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3.2-NVFP4": [
        {
          "name": "DeepSeek-V3.2 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V3.2-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--kv-cache-dtype": "fp8",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--kv-cache-dtype": "fp8",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "meta-llama/Llama-3.3-70B-Instruct": [
        {
          "name": "Llama-3.3-70B",
          "huggingface_id": "meta-llama/Llama-3.3-70B-Instruct",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-FP8": [
        {
          "name": "Llama-3.3-70B (fp8)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-FP8",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-NVFP4": [
        {
          "name": "Llama-3.3-70B (nvfp4)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "meta-llama/Llama-4-Scout-17B-16E-Instruct": [
        {
          "name": "Llama-4-Scout",
          "huggingface_id": "meta-llama/Llama-4-Scout-17B-16E-Instruct",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8": [
        {
          "name": "Llama-4-Scout (fp8)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4": [
        {
          "name": "Llama-4-Scout (nvfp4)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "microsoft/Phi-4-mini-instruct": [
        {
          "name": "Phi-4",
          "huggingface_id": "microsoft/Phi-4-mini-instruct",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "microsoft/Phi-4-mini-reasoning": [
        {
          "name": "Phi-4 (mini_reasoning)",
          "huggingface_id": "microsoft/Phi-4-mini-reasoning",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mini_reasoning"
        }
      ],
      "microsoft/Phi-4-reasoning": [
        {
          "name": "Phi-4 (reasoning)",
          "huggingface_id": "microsoft/Phi-4-reasoning",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "reasoning"
        }
      ],
      "microsoft/Phi-4-multimodal-instruct": [
        {
          "name": "Phi-4 (multimodal)",
          "huggingface_id": "microsoft/Phi-4-multimodal-instruct",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "multimodal"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512": [
        {
          "name": "Mistral-Large-3-675B-Instruct",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4": [
        {
          "name": "Mistral-Large-3-675B-Instruct (nvfp4)",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512-FP8": [
        {
          "name": "Mistral-Large-3-675B-Instruct (fp8)",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512-FP8",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Cosmos3-Nano": [
        {
          "name": "Cosmos3-Nano",
          "huggingface_id": "nvidia/Cosmos3-Nano",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super-Image2Video": [
        {
          "name": "Cosmos3-Super-Image2Video",
          "huggingface_id": "nvidia/Cosmos3-Super-Image2Video",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super-Text2Image": [
        {
          "name": "Cosmos3-Super-Text2Image",
          "huggingface_id": "nvidia/Cosmos3-Super-Text2Image",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super": [
        {
          "name": "Cosmos3-Super",
          "huggingface_id": "nvidia/Cosmos3-Super",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Nano-30B-A3B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Nano-30B-A3B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_FLASHINFER_MOE_BACKEND": "throughput"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_FLASHINFER_MOE_BACKEND": "throughput"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Nano-4B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Nano-4B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (base_bf16)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base_bf16"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-FP8": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-FP8",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-NVFP4-QAD": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-NVFP4-QAD",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-FP8": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-FP8",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-NVFP4": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Base": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (base)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Base",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Japanese": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (japanese)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Japanese",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "japanese"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning (fp8)",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning (nvfp4)",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "openai/gpt-oss-120b": [
        {
          "name": "GPT-OSS",
          "huggingface_id": "openai/gpt-oss-120b",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "amd/gpt-oss-120b-w-mxfp4-a-fp8": [
        {
          "name": "GPT-OSS (amd_fp8)",
          "huggingface_id": "amd/gpt-oss-120b-w-mxfp4-a-fp8",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "amd_fp8"
        }
      ],
      "openai/gpt-oss-20b": [
        {
          "name": "GPT-OSS 20B",
          "huggingface_id": "openai/gpt-oss-20b",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.5V": [
        {
          "name": "GLM-4.5V",
          "huggingface_id": "zai-org/GLM-4.5V",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.5V-FP8": [
        {
          "name": "GLM-4.5V (fp8)",
          "huggingface_id": "zai-org/GLM-4.5V-FP8",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.6V": [
        {
          "name": "GLM-4.6V",
          "huggingface_id": "zai-org/GLM-4.6V",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.6V-FP8": [
        {
          "name": "GLM-4.6V (fp8)",
          "huggingface_id": "zai-org/GLM-4.6V-FP8",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-Image": [
        {
          "name": "GLM-Image",
          "huggingface_id": "zai-org/GLM-Image",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/Glyph": [
        {
          "name": "Glyph",
          "huggingface_id": "zai-org/Glyph",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--limit-mm-per-prompt.video": "0",
                "--reasoning-parser": "glm45",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-26B-A4B-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-26B-A4B-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-31B-IT-NVFP4": [
        {
          "name": "Gemma 4 31B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-31B-IT-NVFP4",
          "accelerators": "h100",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ]
    },
    "h200": {
      "google/diffusiongemma-26B-A4B-it": [
        {
          "name": "DiffusionGemma 26B-A4B IT",
          "huggingface_id": "google/diffusiongemma-26B-A4B-it",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/diffusiongemma-26B-A4B-it-FP8-dynamic": [
        {
          "name": "DiffusionGemma 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/diffusiongemma-26B-A4B-it-FP8-dynamic",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/diffusiongemma-26B-A4B-it-NVFP4": [
        {
          "name": "DiffusionGemma 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/diffusiongemma-26B-A4B-it-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-26B-A4B-it": [
        {
          "name": "Gemma 4 26B-A4B IT",
          "huggingface_id": "google/gemma-4-26B-A4B-it",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "JetBrains/Mellum2-12B-A2.5B-Instruct": [
        {
          "name": "Mellum2-12B-A2.5B-Instruct",
          "huggingface_id": "JetBrains/Mellum2-12B-A2.5B-Instruct",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "JetBrains/Mellum2-12B-A2.5B-Thinking": [
        {
          "name": "Mellum2-12B-A2.5B-Thinking",
          "huggingface_id": "JetBrains/Mellum2-12B-A2.5B-Thinking",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Base": [
        {
          "name": "LFM2.5 1.2B Base",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Base",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Instruct": [
        {
          "name": "LFM2.5 1.2B Instruct",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Instruct",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-JP-202606": [
        {
          "name": "LFM2.5 1.2B JP (202606)",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-JP-202606",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-JP": [
        {
          "name": "LFM2.5 1.2B JP",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-JP",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Thinking": [
        {
          "name": "LFM2.5 1.2B Thinking",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Thinking",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-230M": [
        {
          "name": "LFM2.5 230M",
          "huggingface_id": "LiquidAI/LFM2.5-230M",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-350M": [
        {
          "name": "LFM2.5 350M",
          "huggingface_id": "LiquidAI/LFM2.5-350M",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-8B-A1B": [
        {
          "name": "LFM2.5 8B A1B",
          "huggingface_id": "LiquidAI/LFM2.5-8B-A1B",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-VL-1.6B": [
        {
          "name": "LFM2.5 VL 1.6B",
          "huggingface_id": "LiquidAI/LFM2.5-VL-1.6B",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-VL-450M": [
        {
          "name": "LFM2.5 VL 450M",
          "huggingface_id": "LiquidAI/LFM2.5-VL-450M",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M2.5": [
        {
          "name": "MiniMax-M2.5",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.5",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--compilation-config": "{\"cudagraph_mode\":\"PIECEWISE\"}",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "32768",
                "--kv-cache-dtype": "fp8",
                "--moe-backend": "triton",
                "--attention-backend": "FLASHINFER",
                "--enable-flashinfer-autotune": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "PYTHONNOUSERSITE": "1",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0",
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--compilation-config": "{\"cudagraph_mode\":\"PIECEWISE\"}",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "32768",
                "--kv-cache-dtype": "fp8",
                "--moe-backend": "triton",
                "--attention-backend": "FLASHINFER",
                "--enable-flashinfer-autotune": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "PYTHONNOUSERSITE": "1",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0",
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.5-NVFP4": [
        {
          "name": "MiniMax-M2.5 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.5-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--compilation-config": "{\"cudagraph_mode\":\"PIECEWISE\"}",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "32768",
                "--kv-cache-dtype": "fp8",
                "--moe-backend": "triton",
                "--attention-backend": "FLASHINFER",
                "--enable-flashinfer-autotune": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "PYTHONNOUSERSITE": "1",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0",
                "VLLM_FLOAT32_MATMUL_PRECISION": "high",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--compilation-config": "{\"cudagraph_mode\":\"PIECEWISE\"}",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "32768",
                "--kv-cache-dtype": "fp8",
                "--moe-backend": "triton",
                "--attention-backend": "FLASHINFER",
                "--enable-flashinfer-autotune": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "PYTHONNOUSERSITE": "1",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0",
                "VLLM_FLOAT32_MATMUL_PRECISION": "high",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2.7": [
        {
          "name": "MiniMax-M2.7",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.7",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.7-NVFP4": [
        {
          "name": "MiniMax-M2.7 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.7-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2": [
        {
          "name": "MiniMax-M2",
          "huggingface_id": "MiniMaxAI/MiniMax-M2",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/MiniMax-M2-NVFP4": [
        {
          "name": "MiniMax-M2 (nvfp4)",
          "huggingface_id": "RedHatAI/MiniMax-M2-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M3": [
        {
          "name": "MiniMax-M3",
          "huggingface_id": "MiniMaxAI/MiniMax-M3",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M3-MXFP8": [
        {
          "name": "MiniMax-M3 (mxfp8)",
          "huggingface_id": "MiniMaxAI/MiniMax-M3-MXFP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp8"
        }
      ],
      "nvidia/MiniMax-M3-NVFP4": [
        {
          "name": "MiniMax-M3 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M3-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "amd/MiniMax-M3-MXFP4": [
        {
          "name": "MiniMax-M3 (mxfp4)",
          "huggingface_id": "amd/MiniMax-M3-MXFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp4"
        }
      ],
      "Qwen/Qwen3-235B-A22B-Instruct-2507": [
        {
          "name": "Qwen3-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-Instruct-2507",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-235B-A22B-FP8": [
        {
          "name": "Qwen3-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4": [
        {
          "name": "Qwen3-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_USE_DEEP_GEMM": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_DEEP_GEMM": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_USE_DEEP_GEMM": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_DEEP_GEMM": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_USE_DEEP_GEMM": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_DEEP_GEMM": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B": [
        {
          "name": "Qwen3.5-397B",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Qwen3.5-397B-A17B-NVFP4": [
        {
          "name": "Qwen3.5-397B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.5-397B-A17B-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-397B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B": [
        {
          "name": "Qwen3.6-35B-A3B",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B-FP8": [
        {
          "name": "Qwen3.6-35B-A3B (fp8)",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3.6-35B-A3B-NVFP4": [
        {
          "name": "Qwen3.6-35B-A3B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.6-35B-A3B-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "XiaomiMiMo/MiMo-V2-Flash": [
        {
          "name": "MiMo-V2-Flash",
          "huggingface_id": "XiaomiMiMo/MiMo-V2-Flash",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "XiaomiMiMo/MiMo-V2.5-Pro": [
        {
          "name": "MiMo-V2.5-Pro",
          "huggingface_id": "XiaomiMiMo/MiMo-V2.5-Pro",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "XiaomiMiMo/MiMo-V2.5": [
        {
          "name": "MiMo-V2.5",
          "huggingface_id": "XiaomiMiMo/MiMo-V2.5",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-R1": [
        {
          "name": "DeepSeek-R1",
          "huggingface_id": "deepseek-ai/DeepSeek-R1",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-R1-0528": [
        {
          "name": "DeepSeek-R1 (r1_0528)",
          "huggingface_id": "deepseek-ai/DeepSeek-R1-0528",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "r1_0528"
        }
      ],
      "nvidia/DeepSeek-R1-0528-NVFP4-v2": [
        {
          "name": "DeepSeek-R1 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-R1-0528-NVFP4-v2",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3.1": [
        {
          "name": "DeepSeek-V3.1",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.1",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3.1-NVFP4": [
        {
          "name": "DeepSeek-V3.1 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V3.1-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3.2-Exp": [
        {
          "name": "DeepSeek-V3.2-Exp",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.2-Exp",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-V3.2": [
        {
          "name": "DeepSeek-V3.2",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.2",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3.2-NVFP4": [
        {
          "name": "DeepSeek-V3.2 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V3.2-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--kv-cache-dtype": "fp8",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--kv-cache-dtype": "fp8",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3": [
        {
          "name": "DeepSeek-V3",
          "huggingface_id": "deepseek-ai/DeepSeek-V3",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3-FP4": [
        {
          "name": "DeepSeek-V3 (fp4)",
          "huggingface_id": "nvidia/DeepSeek-V3-FP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Flash": [
        {
          "name": "DeepSeek-V4-Flash",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Flash",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Flash-NVFP4": [
        {
          "name": "DeepSeek-V4-Flash (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Flash-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro": [
        {
          "name": "DeepSeek-V4-Pro",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--gpu-memory-utilization": "0.95",
                "--max-num-seqs": "16",
                "--max-num-batched-tokens": "16384",
                "--no-enable-flashinfer-autotune": "",
                "--compilation-config": "{\"mode\": 0, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--gpu-memory-utilization": "0.95",
                "--max-num-seqs": "16",
                "--max-num-batched-tokens": "16384",
                "--no-enable-flashinfer-autotune": "",
                "--compilation-config": "{\"mode\": 0, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Pro-NVFP4": [
        {
          "name": "DeepSeek-V4-Pro (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Pro-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--gpu-memory-utilization": "0.95",
                "--max-num-seqs": "16",
                "--max-num-batched-tokens": "16384",
                "--no-enable-flashinfer-autotune": "",
                "--compilation-config": "{\"mode\": 0, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--gpu-memory-utilization": "0.95",
                "--max-num-seqs": "16",
                "--max-num-batched-tokens": "16384",
                "--no-enable-flashinfer-autotune": "",
                "--compilation-config": "{\"mode\": 0, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro-DSpark": [
        {
          "name": "DeepSeek-V4-Pro (dspark)",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro-DSpark",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--gpu-memory-utilization": "0.95",
                "--max-num-seqs": "16",
                "--max-num-batched-tokens": "16384",
                "--no-enable-flashinfer-autotune": "",
                "--compilation-config": "{\"mode\": 0, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--gpu-memory-utilization": "0.95",
                "--max-num-seqs": "16",
                "--max-num-batched-tokens": "16384",
                "--no-enable-flashinfer-autotune": "",
                "--compilation-config": "{\"mode\": 0, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "dspark"
        }
      ],
      "inclusionAI/Ling-2.6-flash": [
        {
          "name": "Ling-2.6-flash",
          "huggingface_id": "inclusionAI/Ling-2.6-flash",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ring-1T-FP8": [
        {
          "name": "Ring-1T-FP8",
          "huggingface_id": "inclusionAI/Ring-1T-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S2-Preview": [
        {
          "name": "Intern-S2-Preview",
          "huggingface_id": "internlm/Intern-S2-Preview",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S2-Preview-FP8": [
        {
          "name": "Intern-S2-Preview (fp8)",
          "huggingface_id": "internlm/Intern-S2-Preview-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "meta-llama/Llama-3.3-70B-Instruct": [
        {
          "name": "Llama-3.3-70B",
          "huggingface_id": "meta-llama/Llama-3.3-70B-Instruct",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-FP8": [
        {
          "name": "Llama-3.3-70B (fp8)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-NVFP4": [
        {
          "name": "Llama-3.3-70B (nvfp4)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Ministral-3-8B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (8b)",
          "huggingface_id": "mistralai/Ministral-3-8B-Instruct-2512",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "8b"
        }
      ],
      "mistralai/Ministral-3-3B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (3b)",
          "huggingface_id": "mistralai/Ministral-3-3B-Instruct-2512",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "3b"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512-FP8": [
        {
          "name": "Ministral-3-Instruct (fp8)",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "mistralai/Ministral-3-8B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning",
          "huggingface_id": "mistralai/Ministral-3-8B-Reasoning-2512",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Ministral-3-3B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning (3b)",
          "huggingface_id": "mistralai/Ministral-3-3B-Reasoning-2512",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "3b"
        }
      ],
      "mistralai/Ministral-3-14B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning (14b)",
          "huggingface_id": "mistralai/Ministral-3-14B-Reasoning-2512",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "14b"
        }
      ],
      "moonshotai/Kimi-K2-Instruct": [
        {
          "name": "Kimi-K2-Instruct",
          "huggingface_id": "moonshotai/Kimi-K2-Instruct",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-K2-Thinking": [
        {
          "name": "Kimi-K2-Thinking",
          "huggingface_id": "moonshotai/Kimi-K2-Thinking",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2-Thinking-NVFP4": [
        {
          "name": "Kimi-K2-Thinking (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2-Thinking-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.5": [
        {
          "name": "Kimi-K2.5",
          "huggingface_id": "moonshotai/Kimi-K2.5",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.5-NVFP4": [
        {
          "name": "Kimi-K2.5 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.5-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.6": [
        {
          "name": "Kimi-K2.6",
          "huggingface_id": "moonshotai/Kimi-K2.6",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.6-NVFP4": [
        {
          "name": "Kimi-K2.6 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.6-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.7-Code": [
        {
          "name": "Kimi-K2.7-Code",
          "huggingface_id": "moonshotai/Kimi-K2.7-Code",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Nano": [
        {
          "name": "Cosmos3-Nano",
          "huggingface_id": "nvidia/Cosmos3-Nano",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super-Image2Video": [
        {
          "name": "Cosmos3-Super-Image2Video",
          "huggingface_id": "nvidia/Cosmos3-Super-Image2Video",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super-Text2Image": [
        {
          "name": "Cosmos3-Super-Text2Image",
          "huggingface_id": "nvidia/Cosmos3-Super-Text2Image",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super": [
        {
          "name": "Cosmos3-Super",
          "huggingface_id": "nvidia/Cosmos3-Super",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Nano-30B-A3B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Nano-30B-A3B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_FLASHINFER_MOE_BACKEND": "throughput"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_FLASHINFER_MOE_BACKEND": "throughput"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Nano-4B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Nano-4B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (base_bf16)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base_bf16"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-FP8": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-NVFP4": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Base": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (base)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Base",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Japanese": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (japanese)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Japanese",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "japanese"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning (fp8)",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning (nvfp4)",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "openai/gpt-oss-120b": [
        {
          "name": "GPT-OSS",
          "huggingface_id": "openai/gpt-oss-120b",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "amd/gpt-oss-120b-w-mxfp4-a-fp8": [
        {
          "name": "GPT-OSS (amd_fp8)",
          "huggingface_id": "amd/gpt-oss-120b-w-mxfp4-a-fp8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "amd_fp8"
        }
      ],
      "openai/gpt-oss-20b": [
        {
          "name": "GPT-OSS 20B",
          "huggingface_id": "openai/gpt-oss-20b",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-XS-2.1": [
        {
          "name": "Laguna XS-2.1",
          "huggingface_id": "poolside/Laguna-XS-2.1",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-XS-2.1-FP8": [
        {
          "name": "Laguna XS-2.1 (fp8)",
          "huggingface_id": "poolside/Laguna-XS-2.1-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "poolside/Laguna-XS-2.1-NVFP4": [
        {
          "name": "Laguna XS-2.1 (nvfp4)",
          "huggingface_id": "poolside/Laguna-XS-2.1-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "poolside/Laguna-XS-2.1-INT4": [
        {
          "name": "Laguna XS-2.1 (int4)",
          "huggingface_id": "poolside/Laguna-XS-2.1-INT4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "poolside/Laguna-XS.2": [
        {
          "name": "Laguna XS.2",
          "huggingface_id": "poolside/Laguna-XS.2",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-XS.2-FP8": [
        {
          "name": "Laguna XS.2 (fp8)",
          "huggingface_id": "poolside/Laguna-XS.2-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "poolside/Laguna-XS.2-NVFP4": [
        {
          "name": "Laguna XS.2 (nvfp4)",
          "huggingface_id": "poolside/Laguna-XS.2-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "poolside/Laguna-XS.2-INT4": [
        {
          "name": "Laguna XS.2 (int4)",
          "huggingface_id": "poolside/Laguna-XS.2-INT4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "stepfun-ai/Step-3.5-Flash": [
        {
          "name": "Step-3.5-Flash",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-FP8": [
        {
          "name": "Step-3.5-Flash (fp8)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-INT4": [
        {
          "name": "Step-3.5-Flash (int4)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-INT4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-INT8": [
        {
          "name": "Step-3.5-Flash (int8)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-INT8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int8"
        }
      ],
      "stepfun-ai/Step-3.7-Flash": [
        {
          "name": "Step-3.7-Flash",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stepfun-ai/Step-3.7-Flash-FP8": [
        {
          "name": "Step-3.7-Flash (fp8)",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "stepfun-ai/Step-3.7-Flash-NVFP4": [
        {
          "name": "Step-3.7-Flash (nvfp4)",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--quantization": "modelopt",
                "--kv-cache-dtype": "fp8",
                "--gpu-memory-utilization": "0.9",
                "--async-scheduling": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--quantization": "modelopt",
                "--kv-cache-dtype": "fp8",
                "--gpu-memory-utilization": "0.9",
                "--async-scheduling": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "tencent/Hy3-preview": [
        {
          "name": "Hy3-preview",
          "huggingface_id": "tencent/Hy3-preview",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "thinkingmachines/Inkling-NVFP4": [
        {
          "name": "TML Inkling",
          "huggingface_id": "thinkingmachines/Inkling-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "thinkingmachines/Inkling": [
        {
          "name": "TML Inkling (bf16)",
          "huggingface_id": "thinkingmachines/Inkling",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "zai-org/GLM-4.5": [
        {
          "name": "GLM-4.5",
          "huggingface_id": "zai-org/GLM-4.5",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.5-FP8": [
        {
          "name": "GLM-4.5 (fp8)",
          "huggingface_id": "zai-org/GLM-4.5-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.5V": [
        {
          "name": "GLM-4.5V",
          "huggingface_id": "zai-org/GLM-4.5V",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.5V-FP8": [
        {
          "name": "GLM-4.5V (fp8)",
          "huggingface_id": "zai-org/GLM-4.5V-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.6": [
        {
          "name": "GLM-4.6",
          "huggingface_id": "zai-org/GLM-4.6",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.6-FP8": [
        {
          "name": "GLM-4.6 (fp8)",
          "huggingface_id": "zai-org/GLM-4.6-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.6V": [
        {
          "name": "GLM-4.6V",
          "huggingface_id": "zai-org/GLM-4.6V",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.6V-FP8": [
        {
          "name": "GLM-4.6V (fp8)",
          "huggingface_id": "zai-org/GLM-4.6V-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.7": [
        {
          "name": "GLM-4.7",
          "huggingface_id": "zai-org/GLM-4.7",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.7-FP8": [
        {
          "name": "GLM-4.7 (fp8)",
          "huggingface_id": "zai-org/GLM-4.7-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-4.7-NVFP4": [
        {
          "name": "GLM-4.7 (nvfp4)",
          "huggingface_id": "nvidia/GLM-4.7-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.1": [
        {
          "name": "GLM-5.1",
          "huggingface_id": "zai-org/GLM-5.1",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-5.1-FP8": [
        {
          "name": "GLM-5.1 (fp8)",
          "huggingface_id": "zai-org/GLM-5.1-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-5.1-NVFP4": [
        {
          "name": "GLM-5.1 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5.1-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5": [
        {
          "name": "GLM-5",
          "huggingface_id": "zai-org/GLM-5",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-5-FP8": [
        {
          "name": "GLM-5 (fp8)",
          "huggingface_id": "zai-org/GLM-5-FP8",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-5-NVFP4": [
        {
          "name": "GLM-5 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/Gemma-4-26B-A4B-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-26B-A4B-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-31B-IT-NVFP4": [
        {
          "name": "Gemma 4 31B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-31B-IT-NVFP4",
          "accelerators": "h200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ]
    },
    "b200": {
      "LiquidAI/LFM2.5-1.2B-Base": [
        {
          "name": "LFM2.5 1.2B Base",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Base",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {
                "VLLM_USE_OINK_OPS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Instruct": [
        {
          "name": "LFM2.5 1.2B Instruct",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Instruct",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {
                "VLLM_USE_OINK_OPS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-JP-202606": [
        {
          "name": "LFM2.5 1.2B JP (202606)",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-JP-202606",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {
                "VLLM_USE_OINK_OPS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-JP": [
        {
          "name": "LFM2.5 1.2B JP",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-JP",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {
                "VLLM_USE_OINK_OPS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Thinking": [
        {
          "name": "LFM2.5 1.2B Thinking",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Thinking",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {
                "VLLM_USE_OINK_OPS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-230M": [
        {
          "name": "LFM2.5 230M",
          "huggingface_id": "LiquidAI/LFM2.5-230M",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-350M": [
        {
          "name": "LFM2.5 350M",
          "huggingface_id": "LiquidAI/LFM2.5-350M",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {
                "VLLM_USE_OINK_OPS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-8B-A1B": [
        {
          "name": "LFM2.5 8B A1B",
          "huggingface_id": "LiquidAI/LFM2.5-8B-A1B",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {
                "VLLM_USE_OINK_OPS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_OINK_OPS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-VL-1.6B": [
        {
          "name": "LFM2.5 VL 1.6B",
          "huggingface_id": "LiquidAI/LFM2.5-VL-1.6B",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {
                "VLLM_USE_OINK_OPS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-VL-450M": [
        {
          "name": "LFM2.5 VL 450M",
          "huggingface_id": "LiquidAI/LFM2.5-VL-450M",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {
                "VLLM_USE_OINK_OPS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M3": [
        {
          "name": "MiniMax-M3",
          "huggingface_id": "MiniMaxAI/MiniMax-M3",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M3-MXFP8": [
        {
          "name": "MiniMax-M3 (mxfp8)",
          "huggingface_id": "MiniMaxAI/MiniMax-M3-MXFP8",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp8"
        }
      ],
      "nvidia/MiniMax-M3-NVFP4": [
        {
          "name": "MiniMax-M3 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M3-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "amd/MiniMax-M3-MXFP4": [
        {
          "name": "MiniMax-M3 (mxfp4)",
          "huggingface_id": "amd/MiniMax-M3-MXFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp4"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-R1": [
        {
          "name": "DeepSeek-R1",
          "huggingface_id": "deepseek-ai/DeepSeek-R1",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-R1-0528": [
        {
          "name": "DeepSeek-R1 (r1_0528)",
          "huggingface_id": "deepseek-ai/DeepSeek-R1-0528",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "r1_0528"
        }
      ],
      "nvidia/DeepSeek-R1-0528-NVFP4-v2": [
        {
          "name": "DeepSeek-R1 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-R1-0528-NVFP4-v2",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3": [
        {
          "name": "DeepSeek-V3",
          "huggingface_id": "deepseek-ai/DeepSeek-V3",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3-FP4": [
        {
          "name": "DeepSeek-V3 (fp4)",
          "huggingface_id": "nvidia/DeepSeek-V3-FP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Flash": [
        {
          "name": "DeepSeek-V4-Flash",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Flash",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Flash-NVFP4": [
        {
          "name": "DeepSeek-V4-Flash (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Flash-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro": [
        {
          "name": "DeepSeek-V4-Pro",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Pro-NVFP4": [
        {
          "name": "DeepSeek-V4-Pro (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Pro-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro-DSpark": [
        {
          "name": "DeepSeek-V4-Pro (dspark)",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro-DSpark",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "dspark"
        }
      ],
      "meta-llama/Llama-3.3-70B-Instruct": [
        {
          "name": "Llama-3.3-70B",
          "huggingface_id": "meta-llama/Llama-3.3-70B-Instruct",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"fuse_attn_quant\":true,\"eliminate_noops\":true}}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-FP8": [
        {
          "name": "Llama-3.3-70B (fp8)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-FP8",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"fuse_attn_quant\":true,\"eliminate_noops\":true}}",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-NVFP4": [
        {
          "name": "Llama-3.3-70B (nvfp4)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"fuse_attn_quant\":true,\"eliminate_noops\":true}}",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "meta-llama/Llama-4-Scout-17B-16E-Instruct": [
        {
          "name": "Llama-4-Scout",
          "huggingface_id": "meta-llama/Llama-4-Scout-17B-16E-Instruct",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8": [
        {
          "name": "Llama-4-Scout (fp8)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4": [
        {
          "name": "Llama-4-Scout (nvfp4)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512": [
        {
          "name": "Mistral-Large-3-675B-Instruct",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4": [
        {
          "name": "Mistral-Large-3-675B-Instruct (nvfp4)",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512-FP8": [
        {
          "name": "Mistral-Large-3-675B-Instruct (fp8)",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512-FP8",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Nano-4B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Nano-4B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (base_bf16)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base_bf16"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Ultra-550B-A55B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}"
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4": [
        {
          "name": "NVIDIA Nemotron-3-Ultra-550B-A55B (fp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}"
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-FP8": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-FP8",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-NVFP4-QAD": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-NVFP4-QAD",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-FP8": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-FP8",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-NVFP4": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Base": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (base)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Base",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Japanese": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (japanese)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Japanese",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "japanese"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning (fp8)",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning (nvfp4)",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "openai/gpt-oss-120b": [
        {
          "name": "GPT-OSS",
          "huggingface_id": "openai/gpt-oss-120b",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_MXFP4_MXFP8": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_MXFP4_MXFP8": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "amd/gpt-oss-120b-w-mxfp4-a-fp8": [
        {
          "name": "GPT-OSS (amd_fp8)",
          "huggingface_id": "amd/gpt-oss-120b-w-mxfp4-a-fp8",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_MXFP4_MXFP8": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_MXFP4_MXFP8": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "amd_fp8"
        }
      ],
      "openai/gpt-oss-20b": [
        {
          "name": "GPT-OSS 20B",
          "huggingface_id": "openai/gpt-oss-20b",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_MXFP4_MXFP8": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_MXFP4_MXFP8": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stepfun-ai/Step-3.5-Flash": [
        {
          "name": "Step-3.5-Flash",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-FP8": [
        {
          "name": "Step-3.5-Flash (fp8)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-FP8",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-INT4": [
        {
          "name": "Step-3.5-Flash (int4)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-INT4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-INT8": [
        {
          "name": "Step-3.5-Flash (int8)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-INT8",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int8"
        }
      ],
      "stepfun-ai/Step-3.7-Flash": [
        {
          "name": "Step-3.7-Flash",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stepfun-ai/Step-3.7-Flash-FP8": [
        {
          "name": "Step-3.7-Flash (fp8)",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash-FP8",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "stepfun-ai/Step-3.7-Flash-NVFP4": [
        {
          "name": "Step-3.7-Flash (nvfp4)",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--quantization": "modelopt",
                "--kv-cache-dtype": "fp8",
                "--gpu-memory-utilization": "0.9",
                "--async-scheduling": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--quantization": "modelopt",
                "--kv-cache-dtype": "fp8",
                "--gpu-memory-utilization": "0.9",
                "--async-scheduling": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "thinkingmachines/Inkling-NVFP4": [
        {
          "name": "TML Inkling",
          "huggingface_id": "thinkingmachines/Inkling-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "thinkingmachines/Inkling": [
        {
          "name": "TML Inkling (bf16)",
          "huggingface_id": "thinkingmachines/Inkling",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "zai-org/GLM-5.2-FP8": [
        {
          "name": "GLM-5.2",
          "huggingface_id": "zai-org/GLM-5.2-FP8",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/GLM-5.2-NVFP4": [
        {
          "name": "GLM-5.2 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5.2-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.2": [
        {
          "name": "GLM-5.2 (bf16)",
          "huggingface_id": "zai-org/GLM-5.2",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "nvidia/Gemma-4-26B-A4B-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-26B-A4B-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-31B-IT-NVFP4": [
        {
          "name": "Gemma 4 31B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-31B-IT-NVFP4",
          "accelerators": "b200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ]
    },
    "gb200": {
      "Qwen/Qwen3.5-397B-A17B": [
        {
          "name": "Qwen3.5-397B",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Qwen3.5-397B-A17B-NVFP4": [
        {
          "name": "Qwen3.5-397B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.5-397B-A17B-NVFP4",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-397B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "deepseek-ai/DeepSeek-R1": [
        {
          "name": "DeepSeek-R1",
          "huggingface_id": "deepseek-ai/DeepSeek-R1",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-R1-0528": [
        {
          "name": "DeepSeek-R1 (r1_0528)",
          "huggingface_id": "deepseek-ai/DeepSeek-R1-0528",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "r1_0528"
        }
      ],
      "nvidia/DeepSeek-R1-0528-NVFP4-v2": [
        {
          "name": "DeepSeek-R1 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-R1-0528-NVFP4-v2",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3": [
        {
          "name": "DeepSeek-V3",
          "huggingface_id": "deepseek-ai/DeepSeek-V3",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3-FP4": [
        {
          "name": "DeepSeek-V3 (fp4)",
          "huggingface_id": "nvidia/DeepSeek-V3-FP4",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Flash": [
        {
          "name": "DeepSeek-V4-Flash",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Flash",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Flash-NVFP4": [
        {
          "name": "DeepSeek-V4-Flash (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Flash-NVFP4",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro": [
        {
          "name": "DeepSeek-V4-Pro",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Pro-NVFP4": [
        {
          "name": "DeepSeek-V4-Pro (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Pro-NVFP4",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro-DSpark": [
        {
          "name": "DeepSeek-V4-Pro (dspark)",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro-DSpark",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "dspark"
        }
      ],
      "internlm/Intern-S2-Preview": [
        {
          "name": "Intern-S2-Preview",
          "huggingface_id": "internlm/Intern-S2-Preview",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S2-Preview-FP8": [
        {
          "name": "Intern-S2-Preview (fp8)",
          "huggingface_id": "internlm/Intern-S2-Preview-FP8",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "meta-llama/Llama-3.3-70B-Instruct": [
        {
          "name": "Llama-3.3-70B",
          "huggingface_id": "meta-llama/Llama-3.3-70B-Instruct",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"fuse_attn_quant\":true,\"eliminate_noops\":true}}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-FP8": [
        {
          "name": "Llama-3.3-70B (fp8)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-FP8",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"fuse_attn_quant\":true,\"eliminate_noops\":true}}",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-NVFP4": [
        {
          "name": "Llama-3.3-70B (nvfp4)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-NVFP4",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"fuse_attn_quant\":true,\"eliminate_noops\":true}}",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "meta-llama/Llama-4-Scout-17B-16E-Instruct": [
        {
          "name": "Llama-4-Scout",
          "huggingface_id": "meta-llama/Llama-4-Scout-17B-16E-Instruct",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8": [
        {
          "name": "Llama-4-Scout (fp8)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4": [
        {
          "name": "Llama-4-Scout (nvfp4)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--async-scheduling": "",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"pass_config\":{\"fuse_allreduce_rms\":true,\"eliminate_noops\":true}}",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "mistralai/Ministral-3-8B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning",
          "huggingface_id": "mistralai/Ministral-3-8B-Reasoning-2512",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Ministral-3-3B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning (3b)",
          "huggingface_id": "mistralai/Ministral-3-3B-Reasoning-2512",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "3b"
        }
      ],
      "mistralai/Ministral-3-14B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning (14b)",
          "huggingface_id": "mistralai/Ministral-3-14B-Reasoning-2512",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "14b"
        }
      ],
      "moonshotai/Kimi-K2.5": [
        {
          "name": "Kimi-K2.5",
          "huggingface_id": "moonshotai/Kimi-K2.5",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-config.use_trtllm_ragged_deepseek_prefill=True": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-config.use_trtllm_ragged_deepseek_prefill=True": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.5-NVFP4": [
        {
          "name": "Kimi-K2.5 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.5-NVFP4",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-config.use_trtllm_ragged_deepseek_prefill=True": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-config.use_trtllm_ragged_deepseek_prefill=True": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.6": [
        {
          "name": "Kimi-K2.6",
          "huggingface_id": "moonshotai/Kimi-K2.6",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-config": "{\"mla_prefill_backend\":\"TRTLLM_RAGGED\",\"use_prefill_query_quantization\":true}",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-config": "{\"mla_prefill_backend\":\"TRTLLM_RAGGED\",\"use_prefill_query_quantization\":true}",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.6-NVFP4": [
        {
          "name": "Kimi-K2.6 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.6-NVFP4",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-config": "{\"mla_prefill_backend\":\"TRTLLM_RAGGED\",\"use_prefill_query_quantization\":true}",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-config": "{\"mla_prefill_backend\":\"TRTLLM_RAGGED\",\"use_prefill_query_quantization\":true}",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/Cosmos3-Super-Text2Image": [
        {
          "name": "Cosmos3-Super-Text2Image",
          "huggingface_id": "nvidia/Cosmos3-Super-Text2Image",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "thinkingmachines/Inkling-NVFP4": [
        {
          "name": "TML Inkling",
          "huggingface_id": "thinkingmachines/Inkling-NVFP4",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "thinkingmachines/Inkling": [
        {
          "name": "TML Inkling (bf16)",
          "huggingface_id": "thinkingmachines/Inkling",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "nvidia/Gemma-4-26B-A4B-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-26B-A4B-NVFP4",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-31B-IT-NVFP4": [
        {
          "name": "Gemma 4 31B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-31B-IT-NVFP4",
          "accelerators": "gb200",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ]
    },
    "b300": {
      "deepseek-ai/DeepSeek-V4-Flash": [
        {
          "name": "DeepSeek-V4-Flash",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Flash",
          "accelerators": "b300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Flash-NVFP4": [
        {
          "name": "DeepSeek-V4-Flash (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Flash-NVFP4",
          "accelerators": "b300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro": [
        {
          "name": "DeepSeek-V4-Pro",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro",
          "accelerators": "b300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Pro-NVFP4": [
        {
          "name": "DeepSeek-V4-Pro (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Pro-NVFP4",
          "accelerators": "b300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro-DSpark": [
        {
          "name": "DeepSeek-V4-Pro (dspark)",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro-DSpark",
          "accelerators": "b300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "dspark"
        }
      ],
      "inclusionAI/Ling-2.6-1T": [
        {
          "name": "Ling-2.6-1T",
          "huggingface_id": "inclusionAI/Ling-2.6-1T",
          "accelerators": "b300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-K2.6": [
        {
          "name": "Kimi-K2.6",
          "huggingface_id": "moonshotai/Kimi-K2.6",
          "accelerators": "b300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-config": "{\"mla_prefill_backend\":\"TRTLLM_RAGGED\",\"use_prefill_query_quantization\":true}",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-config": "{\"mla_prefill_backend\":\"TRTLLM_RAGGED\",\"use_prefill_query_quantization\":true}",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.6-NVFP4": [
        {
          "name": "Kimi-K2.6 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.6-NVFP4",
          "accelerators": "b300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-config": "{\"mla_prefill_backend\":\"TRTLLM_RAGGED\",\"use_prefill_query_quantization\":true}",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-config": "{\"mla_prefill_backend\":\"TRTLLM_RAGGED\",\"use_prefill_query_quantization\":true}",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.2-FP8": [
        {
          "name": "GLM-5.2",
          "huggingface_id": "zai-org/GLM-5.2-FP8",
          "accelerators": "b300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/GLM-5.2-NVFP4": [
        {
          "name": "GLM-5.2 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5.2-NVFP4",
          "accelerators": "b300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.2": [
        {
          "name": "GLM-5.2 (bf16)",
          "huggingface_id": "zai-org/GLM-5.2",
          "accelerators": "b300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "nvidia/Gemma-4-26B-A4B-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-26B-A4B-NVFP4",
          "accelerators": "b300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-31B-IT-NVFP4": [
        {
          "name": "Gemma 4 31B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-31B-IT-NVFP4",
          "accelerators": "b300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ]
    },
    "gb300": {
      "deepseek-ai/DeepSeek-V4-Flash": [
        {
          "name": "DeepSeek-V4-Flash",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Flash",
          "accelerators": "gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Flash-NVFP4": [
        {
          "name": "DeepSeek-V4-Flash (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Flash-NVFP4",
          "accelerators": "gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro": [
        {
          "name": "DeepSeek-V4-Pro",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro",
          "accelerators": "gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Pro-NVFP4": [
        {
          "name": "DeepSeek-V4-Pro (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Pro-NVFP4",
          "accelerators": "gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro-DSpark": [
        {
          "name": "DeepSeek-V4-Pro (dspark)",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro-DSpark",
          "accelerators": "gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "dspark"
        }
      ],
      "tencent/Hy3": [
        {
          "name": "Hy3",
          "huggingface_id": "tencent/Hy3",
          "accelerators": "gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_FLASHINFER_ALLREDUCE_BACKEND": "trtllm"
              }
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_FLASHINFER_ALLREDUCE_BACKEND": "trtllm"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3-FP8": [
        {
          "name": "Hy3 (fp8)",
          "huggingface_id": "tencent/Hy3-FP8",
          "accelerators": "gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_FLASHINFER_ALLREDUCE_BACKEND": "trtllm"
              }
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_FLASHINFER_ALLREDUCE_BACKEND": "trtllm"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Gemma-4-26B-A4B-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-26B-A4B-NVFP4",
          "accelerators": "gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-31B-IT-NVFP4": [
        {
          "name": "Gemma 4 31B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-31B-IT-NVFP4",
          "accelerators": "gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ]
    },
    "dgx_station_gb300": {
      "MiniMaxAI/MiniMax-M2.7": [
        {
          "name": "MiniMax-M2.7",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.7",
          "accelerators": "dgx_station_gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.7-NVFP4": [
        {
          "name": "MiniMax-M2.7 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.7-NVFP4",
          "accelerators": "dgx_station_gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Flash": [
        {
          "name": "DeepSeek-V4-Flash",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Flash",
          "accelerators": "dgx_station_gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Flash-NVFP4": [
        {
          "name": "DeepSeek-V4-Flash (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Flash-NVFP4",
          "accelerators": "dgx_station_gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--attention_config.use_fp4_indexer_cache=True": "",
                "--moe-backend": "deep_gemm_mega_moe",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
          "accelerators": "dgx_station_gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8",
          "accelerators": "dgx_station_gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4",
          "accelerators": "dgx_station_gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (base_bf16)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16",
          "accelerators": "dgx_station_gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base_bf16"
        }
      ],
      "stepfun-ai/Step-3.7-Flash": [
        {
          "name": "Step-3.7-Flash",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash",
          "accelerators": "dgx_station_gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stepfun-ai/Step-3.7-Flash-FP8": [
        {
          "name": "Step-3.7-Flash (fp8)",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash-FP8",
          "accelerators": "dgx_station_gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "stepfun-ai/Step-3.7-Flash-NVFP4": [
        {
          "name": "Step-3.7-Flash (nvfp4)",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash-NVFP4",
          "accelerators": "dgx_station_gb300",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--quantization": "modelopt",
                "--kv-cache-dtype": "fp8",
                "--gpu-memory-utilization": "0.9",
                "--async-scheduling": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--quantization": "modelopt",
                "--kv-cache-dtype": "fp8",
                "--gpu-memory-utilization": "0.9",
                "--async-scheduling": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ]
    },
    "dgx_spark_gb10": {
      "Qwen/Qwen3.6-35B-A3B": [
        {
          "name": "Qwen3.6-35B-A3B",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B",
          "accelerators": "dgx_spark_gb10",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B-FP8": [
        {
          "name": "Qwen3.6-35B-A3B (fp8)",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B-FP8",
          "accelerators": "dgx_spark_gb10",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3.6-35B-A3B-NVFP4": [
        {
          "name": "Qwen3.6-35B-A3B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.6-35B-A3B-NVFP4",
          "accelerators": "dgx_spark_gb10",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
          "accelerators": "dgx_spark_gb10",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8",
          "accelerators": "dgx_spark_gb10",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4",
          "accelerators": "dgx_spark_gb10",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (base_bf16)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16",
          "accelerators": "dgx_spark_gb10",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base_bf16"
        }
      ]
    },
    "mi300x": {
      "google/gemma-4-26B-A4B-it": [
        {
          "name": "Gemma 4 26B-A4B IT",
          "huggingface_id": "google/gemma-4-26B-A4B-it",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it": [
        {
          "name": "Gemma 4 31B IT",
          "huggingface_id": "google/gemma-4-31B-it",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-31B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 31B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-31B-it-FP8-dynamic",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/gemma-4-31B-it-NVFP4": [
        {
          "name": "Gemma 4 31B IT (nvfp4)",
          "huggingface_id": "nvidia/gemma-4-31B-it-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 31B IT (w4a16)",
          "huggingface_id": "google/gemma-4-31B-it-qat-w4a16-ct",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it": [
        {
          "name": "Gemma 4 E2B IT",
          "huggingface_id": "google/gemma-4-E2B-it",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E2B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E2B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E2B-it-FP8-dynamic",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E2B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E2B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-w4a16-ct",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E2B IT (mobile)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-mobile-ct",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "google/gemma-4-E4B-it": [
        {
          "name": "Gemma 4 E4B IT",
          "huggingface_id": "google/gemma-4-E4B-it",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E4B-it-FP8-dynamic",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E4B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E4B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-w4a16-ct",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E4B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E4B IT (mobile)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-mobile-ct",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "MiniMaxAI/MiniMax-M2.1": [
        {
          "name": "MiniMax-M2.1",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.1",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M2.5": [
        {
          "name": "MiniMax-M2.5",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.5",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.5-NVFP4": [
        {
          "name": "MiniMax-M2.5 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.5-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2.7": [
        {
          "name": "MiniMax-M2.7",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.7",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.7-NVFP4": [
        {
          "name": "MiniMax-M2.7 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.7-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2": [
        {
          "name": "MiniMax-M2",
          "huggingface_id": "MiniMaxAI/MiniMax-M2",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/MiniMax-M2-NVFP4": [
        {
          "name": "MiniMax-M2 (nvfp4)",
          "huggingface_id": "RedHatAI/MiniMax-M2-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M3": [
        {
          "name": "MiniMax-M3",
          "huggingface_id": "MiniMaxAI/MiniMax-M3",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M3-MXFP8": [
        {
          "name": "MiniMax-M3 (mxfp8)",
          "huggingface_id": "MiniMaxAI/MiniMax-M3-MXFP8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp8"
        }
      ],
      "nvidia/MiniMax-M3-NVFP4": [
        {
          "name": "MiniMax-M3 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M3-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0",
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0",
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "amd/MiniMax-M3-MXFP4": [
        {
          "name": "MiniMax-M3 (mxfp4)",
          "huggingface_id": "amd/MiniMax-M3-MXFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp4"
        }
      ],
      "OpenGVLab/InternVL3_5-8B": [
        {
          "name": "InternVL3.5",
          "huggingface_id": "OpenGVLab/InternVL3_5-8B",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen2.5-VL-72B-Instruct": [
        {
          "name": "Qwen2.5-VL-72B-Instruct",
          "huggingface_id": "Qwen/Qwen2.5-VL-72B-Instruct",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--limit-mm-per-prompt": "{\"image\":2,\"video\":0}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen2.5-VL-72B-Instruct-AWQ": [
        {
          "name": "Qwen2.5-VL-72B-Instruct (awq)",
          "huggingface_id": "Qwen/Qwen2.5-VL-72B-Instruct-AWQ",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--limit-mm-per-prompt": "{\"image\":2,\"video\":0}",
                "--quantization": "awq"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen3-235B-A22B-Instruct-2507": [
        {
          "name": "Qwen3-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-Instruct-2507",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-235B-A22B-FP8": [
        {
          "name": "Qwen3-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-FP8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4": [
        {
          "name": "Qwen3-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B": [
        {
          "name": "Qwen3.5-397B",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Qwen3.5-397B-A17B-NVFP4": [
        {
          "name": "Qwen3.5-397B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.5-397B-A17B-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-397B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B": [
        {
          "name": "Qwen3.6-35B-A3B",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B-FP8": [
        {
          "name": "Qwen3.6-35B-A3B (fp8)",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B-FP8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3.6-35B-A3B-NVFP4": [
        {
          "name": "Qwen3.6-35B-A3B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.6-35B-A3B-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3Guard-Gen-8B": [
        {
          "name": "Qwen3Guard-Gen-8B",
          "huggingface_id": "Qwen/Qwen3Guard-Gen-8B",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3Guard-Gen-4B": [
        {
          "name": "Qwen3Guard-Gen-8B (small_4b)",
          "huggingface_id": "Qwen/Qwen3Guard-Gen-4B",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "small_4b"
        }
      ],
      "Qwen/Qwen3Guard-Gen-0.6B": [
        {
          "name": "Qwen3Guard-Gen-8B (tiny_0_6b)",
          "huggingface_id": "Qwen/Qwen3Guard-Gen-0.6B",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "tiny_0_6b"
        }
      ],
      "XiaomiMiMo/MiMo-V2-Flash": [
        {
          "name": "MiMo-V2-Flash",
          "huggingface_id": "XiaomiMiMo/MiMo-V2-Flash",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "baidu/ERNIE-4.5-VL-28B-A3B-PT": [
        {
          "name": "ERNIE-4.5-VL",
          "huggingface_id": "baidu/ERNIE-4.5-VL-28B-A3B-PT",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "baidu/ERNIE-4.5-VL-424B-A47B-PT": [
        {
          "name": "ERNIE-4.5-VL (424b)",
          "huggingface_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "424b"
        },
        {
          "name": "ERNIE-4.5-VL (424b_fp8)",
          "huggingface_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--quantization": "fp8",
                "--cpu-offload-gb": "50",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--quantization": "fp8",
                "--cpu-offload-gb": "50",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "424b_fp8"
        }
      ],
      "deepseek-ai/DeepSeek-V3": [
        {
          "name": "DeepSeek-V3",
          "huggingface_id": "deepseek-ai/DeepSeek-V3",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "ghcr.io/embeddedllm/vllm-rocm:ubuntu22.04_rocm6.4.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default",
          "extra_envs": {
            "VLLM_RPC_TIMEOUT": "1800000",
            "SAFETENSORS_FAST_GPU": "1"
          }
        }
      ],
      "nvidia/DeepSeek-V3-FP4": [
        {
          "name": "DeepSeek-V3 (fp4)",
          "huggingface_id": "nvidia/DeepSeek-V3-FP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Flash": [
        {
          "name": "DeepSeek-V4-Flash",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Flash",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Flash-NVFP4": [
        {
          "name": "DeepSeek-V4-Flash (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Flash-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro": [
        {
          "name": "DeepSeek-V4-Pro",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Pro-NVFP4": [
        {
          "name": "DeepSeek-V4-Pro (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Pro-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro-DSpark": [
        {
          "name": "DeepSeek-V4-Pro (dspark)",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro-DSpark",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "dspark"
        }
      ],
      "inclusionAI/Ling-2.6-1T": [
        {
          "name": "Ling-2.6-1T",
          "huggingface_id": "inclusionAI/Ling-2.6-1T",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ling-2.6-flash": [
        {
          "name": "Ling-2.6-flash",
          "huggingface_id": "inclusionAI/Ling-2.6-flash",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ring-1T-FP8": [
        {
          "name": "Ring-1T-FP8",
          "huggingface_id": "inclusionAI/Ring-1T-FP8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S1": [
        {
          "name": "Intern-S1",
          "huggingface_id": "internlm/Intern-S1",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S1-FP8": [
        {
          "name": "Intern-S1 (fp8)",
          "huggingface_id": "internlm/Intern-S1-FP8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Ministral-3-8B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (8b)",
          "huggingface_id": "mistralai/Ministral-3-8B-Instruct-2512",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "8b"
        }
      ],
      "mistralai/Ministral-3-3B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (3b)",
          "huggingface_id": "mistralai/Ministral-3-3B-Instruct-2512",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "3b"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512-FP8": [
        {
          "name": "Ministral-3-Instruct (fp8)",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512-FP8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "mistralai/Ministral-3-8B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning",
          "huggingface_id": "mistralai/Ministral-3-8B-Reasoning-2512",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Ministral-3-3B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning (3b)",
          "huggingface_id": "mistralai/Ministral-3-3B-Reasoning-2512",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "3b"
        }
      ],
      "mistralai/Ministral-3-14B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning (14b)",
          "huggingface_id": "mistralai/Ministral-3-14B-Reasoning-2512",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "14b"
        }
      ],
      "mistralai/Mistral-Medium-3.5-128B": [
        {
          "name": "Mistral-Medium-3.5",
          "huggingface_id": "mistralai/Mistral-Medium-3.5-128B",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Medium-3.5-128B-EAGLE\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-K2-Instruct": [
        {
          "name": "Kimi-K2-Instruct",
          "huggingface_id": "moonshotai/Kimi-K2-Instruct",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2"
              },
              "extra_envs": {
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-K2.5": [
        {
          "name": "Kimi-K2.5",
          "huggingface_id": "moonshotai/Kimi-K2.5",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.5-NVFP4": [
        {
          "name": "Kimi-K2.5 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.5-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.6": [
        {
          "name": "Kimi-K2.6",
          "huggingface_id": "moonshotai/Kimi-K2.6",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.6-NVFP4": [
        {
          "name": "Kimi-K2.6 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.6-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.7-Code": [
        {
          "name": "Kimi-K2.7-Code",
          "huggingface_id": "moonshotai/Kimi-K2.7-Code",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "openai/gpt-oss-120b": [
        {
          "name": "GPT-OSS",
          "huggingface_id": "openai/gpt-oss-120b",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "amd/gpt-oss-120b-w-mxfp4-a-fp8": [
        {
          "name": "GPT-OSS (amd_fp8)",
          "huggingface_id": "amd/gpt-oss-120b-w-mxfp4-a-fp8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "amd_fp8"
        }
      ],
      "openai/gpt-oss-20b": [
        {
          "name": "GPT-OSS 20B",
          "huggingface_id": "openai/gpt-oss-20b",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hunyuan-A13B-Instruct": [
        {
          "name": "Hunyuan-A13B-Instruct",
          "huggingface_id": "tencent/Hunyuan-A13B-Instruct",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hunyuan-A13B-Instruct-FP8": [
        {
          "name": "Hunyuan-A13B-Instruct (fp8)",
          "huggingface_id": "tencent/Hunyuan-A13B-Instruct-FP8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "tencent/Hy3-preview": [
        {
          "name": "Hy3-preview",
          "huggingface_id": "tencent/Hy3-preview",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3": [
        {
          "name": "Hy3",
          "huggingface_id": "tencent/Hy3",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3-FP8": [
        {
          "name": "Hy3 (fp8)",
          "huggingface_id": "tencent/Hy3-FP8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "thinkingmachines/Inkling-NVFP4": [
        {
          "name": "TML Inkling",
          "huggingface_id": "thinkingmachines/Inkling-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "thinkingmachines/Inkling": [
        {
          "name": "TML Inkling (bf16)",
          "huggingface_id": "thinkingmachines/Inkling",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "zai-org/GLM-4.5": [
        {
          "name": "GLM-4.5",
          "huggingface_id": "zai-org/GLM-4.5",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.5-FP8": [
        {
          "name": "GLM-4.5 (fp8)",
          "huggingface_id": "zai-org/GLM-4.5-FP8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.6": [
        {
          "name": "GLM-4.6",
          "huggingface_id": "zai-org/GLM-4.6",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.6-FP8": [
        {
          "name": "GLM-4.6 (fp8)",
          "huggingface_id": "zai-org/GLM-4.6-FP8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.7": [
        {
          "name": "GLM-4.7",
          "huggingface_id": "zai-org/GLM-4.7",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.7-FP8": [
        {
          "name": "GLM-4.7 (fp8)",
          "huggingface_id": "zai-org/GLM-4.7-FP8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-4.7-NVFP4": [
        {
          "name": "GLM-4.7 (nvfp4)",
          "huggingface_id": "nvidia/GLM-4.7-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.2-FP8": [
        {
          "name": "GLM-5.2",
          "huggingface_id": "zai-org/GLM-5.2-FP8",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/GLM-5.2-NVFP4": [
        {
          "name": "GLM-5.2 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5.2-NVFP4",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.2": [
        {
          "name": "GLM-5.2 (bf16)",
          "huggingface_id": "zai-org/GLM-5.2",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "zai-org/Glyph": [
        {
          "name": "Glyph",
          "huggingface_id": "zai-org/Glyph",
          "accelerators": "mi300x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--limit-mm-per-prompt.video": "0",
                "--reasoning-parser": "glm45",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ]
    },
    "mi325x": {
      "google/gemma-4-26B-A4B-it": [
        {
          "name": "Gemma 4 26B-A4B IT",
          "huggingface_id": "google/gemma-4-26B-A4B-it",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it": [
        {
          "name": "Gemma 4 31B IT",
          "huggingface_id": "google/gemma-4-31B-it",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-31B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 31B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-31B-it-FP8-dynamic",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/gemma-4-31B-it-NVFP4": [
        {
          "name": "Gemma 4 31B IT (nvfp4)",
          "huggingface_id": "nvidia/gemma-4-31B-it-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 31B IT (w4a16)",
          "huggingface_id": "google/gemma-4-31B-it-qat-w4a16-ct",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it": [
        {
          "name": "Gemma 4 E2B IT",
          "huggingface_id": "google/gemma-4-E2B-it",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E2B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E2B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E2B-it-FP8-dynamic",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E2B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E2B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-w4a16-ct",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E2B IT (mobile)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-mobile-ct",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "google/gemma-4-E4B-it": [
        {
          "name": "Gemma 4 E4B IT",
          "huggingface_id": "google/gemma-4-E4B-it",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E4B-it-FP8-dynamic",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E4B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E4B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-w4a16-ct",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E4B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E4B IT (mobile)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-mobile-ct",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "MiniMaxAI/MiniMax-M2.1": [
        {
          "name": "MiniMax-M2.1",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.1",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M2.5": [
        {
          "name": "MiniMax-M2.5",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.5",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.5-NVFP4": [
        {
          "name": "MiniMax-M2.5 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.5-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2.7": [
        {
          "name": "MiniMax-M2.7",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.7",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.7-NVFP4": [
        {
          "name": "MiniMax-M2.7 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.7-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2": [
        {
          "name": "MiniMax-M2",
          "huggingface_id": "MiniMaxAI/MiniMax-M2",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/MiniMax-M2-NVFP4": [
        {
          "name": "MiniMax-M2 (nvfp4)",
          "huggingface_id": "RedHatAI/MiniMax-M2-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "OpenGVLab/InternVL3_5-8B": [
        {
          "name": "InternVL3.5",
          "huggingface_id": "OpenGVLab/InternVL3_5-8B",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen2.5-VL-72B-Instruct": [
        {
          "name": "Qwen2.5-VL-72B-Instruct",
          "huggingface_id": "Qwen/Qwen2.5-VL-72B-Instruct",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--limit-mm-per-prompt": "{\"image\":2,\"video\":0}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen2.5-VL-72B-Instruct-AWQ": [
        {
          "name": "Qwen2.5-VL-72B-Instruct (awq)",
          "huggingface_id": "Qwen/Qwen2.5-VL-72B-Instruct-AWQ",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--limit-mm-per-prompt": "{\"image\":2,\"video\":0}",
                "--quantization": "awq"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen3-235B-A22B-Instruct-2507": [
        {
          "name": "Qwen3-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-Instruct-2507",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-235B-A22B-FP8": [
        {
          "name": "Qwen3-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-FP8",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4": [
        {
          "name": "Qwen3-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B": [
        {
          "name": "Qwen3.5-397B",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Qwen3.5-397B-A17B-NVFP4": [
        {
          "name": "Qwen3.5-397B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.5-397B-A17B-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-397B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B": [
        {
          "name": "Qwen3.6-35B-A3B",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B-FP8": [
        {
          "name": "Qwen3.6-35B-A3B (fp8)",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B-FP8",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3.6-35B-A3B-NVFP4": [
        {
          "name": "Qwen3.6-35B-A3B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.6-35B-A3B-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "XiaomiMiMo/MiMo-V2-Flash": [
        {
          "name": "MiMo-V2-Flash",
          "huggingface_id": "XiaomiMiMo/MiMo-V2-Flash",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "baidu/ERNIE-4.5-VL-28B-A3B-PT": [
        {
          "name": "ERNIE-4.5-VL",
          "huggingface_id": "baidu/ERNIE-4.5-VL-28B-A3B-PT",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "baidu/ERNIE-4.5-VL-424B-A47B-PT": [
        {
          "name": "ERNIE-4.5-VL (424b)",
          "huggingface_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "424b"
        },
        {
          "name": "ERNIE-4.5-VL (424b_fp8)",
          "huggingface_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--quantization": "fp8",
                "--cpu-offload-gb": "50",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--quantization": "fp8",
                "--cpu-offload-gb": "50",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "424b_fp8"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Flash": [
        {
          "name": "DeepSeek-V4-Flash",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Flash",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Flash-NVFP4": [
        {
          "name": "DeepSeek-V4-Flash (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Flash-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro": [
        {
          "name": "DeepSeek-V4-Pro",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Pro-NVFP4": [
        {
          "name": "DeepSeek-V4-Pro (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Pro-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro-DSpark": [
        {
          "name": "DeepSeek-V4-Pro (dspark)",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro-DSpark",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "dspark"
        }
      ],
      "inclusionAI/Ring-1T-FP8": [
        {
          "name": "Ring-1T-FP8",
          "huggingface_id": "inclusionAI/Ring-1T-FP8",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S1": [
        {
          "name": "Intern-S1",
          "huggingface_id": "internlm/Intern-S1",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S1-FP8": [
        {
          "name": "Intern-S1 (fp8)",
          "huggingface_id": "internlm/Intern-S1-FP8",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Ministral-3-8B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (8b)",
          "huggingface_id": "mistralai/Ministral-3-8B-Instruct-2512",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "8b"
        }
      ],
      "mistralai/Ministral-3-3B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (3b)",
          "huggingface_id": "mistralai/Ministral-3-3B-Instruct-2512",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "3b"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512-FP8": [
        {
          "name": "Ministral-3-Instruct (fp8)",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512-FP8",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "moonshotai/Kimi-K2-Instruct": [
        {
          "name": "Kimi-K2-Instruct",
          "huggingface_id": "moonshotai/Kimi-K2-Instruct",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2"
              },
              "extra_envs": {
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-K2.5": [
        {
          "name": "Kimi-K2.5",
          "huggingface_id": "moonshotai/Kimi-K2.5",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.5-NVFP4": [
        {
          "name": "Kimi-K2.5 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.5-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.6": [
        {
          "name": "Kimi-K2.6",
          "huggingface_id": "moonshotai/Kimi-K2.6",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.6-NVFP4": [
        {
          "name": "Kimi-K2.6 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.6-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.7-Code": [
        {
          "name": "Kimi-K2.7-Code",
          "huggingface_id": "moonshotai/Kimi-K2.7-Code",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "openai/gpt-oss-120b": [
        {
          "name": "GPT-OSS",
          "huggingface_id": "openai/gpt-oss-120b",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "amd/gpt-oss-120b-w-mxfp4-a-fp8": [
        {
          "name": "GPT-OSS (amd_fp8)",
          "huggingface_id": "amd/gpt-oss-120b-w-mxfp4-a-fp8",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "amd_fp8"
        }
      ],
      "openai/gpt-oss-20b": [
        {
          "name": "GPT-OSS 20B",
          "huggingface_id": "openai/gpt-oss-20b",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hunyuan-A13B-Instruct": [
        {
          "name": "Hunyuan-A13B-Instruct",
          "huggingface_id": "tencent/Hunyuan-A13B-Instruct",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hunyuan-A13B-Instruct-FP8": [
        {
          "name": "Hunyuan-A13B-Instruct (fp8)",
          "huggingface_id": "tencent/Hunyuan-A13B-Instruct-FP8",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "tencent/Hy3-preview": [
        {
          "name": "Hy3-preview",
          "huggingface_id": "tencent/Hy3-preview",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "thinkingmachines/Inkling-NVFP4": [
        {
          "name": "TML Inkling",
          "huggingface_id": "thinkingmachines/Inkling-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "thinkingmachines/Inkling": [
        {
          "name": "TML Inkling (bf16)",
          "huggingface_id": "thinkingmachines/Inkling",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "zai-org/GLM-4.5": [
        {
          "name": "GLM-4.5",
          "huggingface_id": "zai-org/GLM-4.5",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.5-FP8": [
        {
          "name": "GLM-4.5 (fp8)",
          "huggingface_id": "zai-org/GLM-4.5-FP8",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.6": [
        {
          "name": "GLM-4.6",
          "huggingface_id": "zai-org/GLM-4.6",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.6-FP8": [
        {
          "name": "GLM-4.6 (fp8)",
          "huggingface_id": "zai-org/GLM-4.6-FP8",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.7": [
        {
          "name": "GLM-4.7",
          "huggingface_id": "zai-org/GLM-4.7",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.7-FP8": [
        {
          "name": "GLM-4.7 (fp8)",
          "huggingface_id": "zai-org/GLM-4.7-FP8",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-4.7-NVFP4": [
        {
          "name": "GLM-4.7 (nvfp4)",
          "huggingface_id": "nvidia/GLM-4.7-NVFP4",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/Glyph": [
        {
          "name": "Glyph",
          "huggingface_id": "zai-org/Glyph",
          "accelerators": "mi325x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--limit-mm-per-prompt.video": "0",
                "--reasoning-parser": "glm45",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ]
    },
    "mi355x": {
      "google/gemma-4-26B-A4B-it": [
        {
          "name": "Gemma 4 26B-A4B IT",
          "huggingface_id": "google/gemma-4-26B-A4B-it",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it": [
        {
          "name": "Gemma 4 31B IT",
          "huggingface_id": "google/gemma-4-31B-it",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-31B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 31B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-31B-it-FP8-dynamic",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/gemma-4-31B-it-NVFP4": [
        {
          "name": "Gemma 4 31B IT (nvfp4)",
          "huggingface_id": "nvidia/gemma-4-31B-it-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 31B IT (w4a16)",
          "huggingface_id": "google/gemma-4-31B-it-qat-w4a16-ct",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it": [
        {
          "name": "Gemma 4 E2B IT",
          "huggingface_id": "google/gemma-4-E2B-it",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E2B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E2B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E2B-it-FP8-dynamic",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E2B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E2B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-w4a16-ct",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E2B IT (mobile)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-mobile-ct",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "google/gemma-4-E4B-it": [
        {
          "name": "Gemma 4 E4B IT",
          "huggingface_id": "google/gemma-4-E4B-it",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E4B-it-FP8-dynamic",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E4B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E4B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-w4a16-ct",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E4B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E4B IT (mobile)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-mobile-ct",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "MiniMaxAI/MiniMax-M2.1": [
        {
          "name": "MiniMax-M2.1",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.1",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M2.5": [
        {
          "name": "MiniMax-M2.5",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.5",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.5-NVFP4": [
        {
          "name": "MiniMax-M2.5 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.5-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2.7": [
        {
          "name": "MiniMax-M2.7",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.7",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.7-NVFP4": [
        {
          "name": "MiniMax-M2.7 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.7-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2": [
        {
          "name": "MiniMax-M2",
          "huggingface_id": "MiniMaxAI/MiniMax-M2",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/MiniMax-M2-NVFP4": [
        {
          "name": "MiniMax-M2 (nvfp4)",
          "huggingface_id": "RedHatAI/MiniMax-M2-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M3": [
        {
          "name": "MiniMax-M3",
          "huggingface_id": "MiniMaxAI/MiniMax-M3",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M3-MXFP8": [
        {
          "name": "MiniMax-M3 (mxfp8)",
          "huggingface_id": "MiniMaxAI/MiniMax-M3-MXFP8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp8"
        }
      ],
      "nvidia/MiniMax-M3-NVFP4": [
        {
          "name": "MiniMax-M3 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M3-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0",
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0",
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "amd/MiniMax-M3-MXFP4": [
        {
          "name": "MiniMax-M3 (mxfp4)",
          "huggingface_id": "amd/MiniMax-M3-MXFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp4"
        }
      ],
      "OpenGVLab/InternVL3_5-8B": [
        {
          "name": "InternVL3.5",
          "huggingface_id": "OpenGVLab/InternVL3_5-8B",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen2.5-VL-72B-Instruct": [
        {
          "name": "Qwen2.5-VL-72B-Instruct",
          "huggingface_id": "Qwen/Qwen2.5-VL-72B-Instruct",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--limit-mm-per-prompt": "{\"image\":2,\"video\":0}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen2.5-VL-72B-Instruct-AWQ": [
        {
          "name": "Qwen2.5-VL-72B-Instruct (awq)",
          "huggingface_id": "Qwen/Qwen2.5-VL-72B-Instruct-AWQ",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--limit-mm-per-prompt": "{\"image\":2,\"video\":0}",
                "--quantization": "awq"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen3-235B-A22B-Instruct-2507": [
        {
          "name": "Qwen3-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-Instruct-2507",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-235B-A22B-FP8": [
        {
          "name": "Qwen3-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-FP8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4": [
        {
          "name": "Qwen3-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B": [
        {
          "name": "Qwen3.5-397B",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Qwen3.5-397B-A17B-NVFP4": [
        {
          "name": "Qwen3.5-397B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.5-397B-A17B-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-397B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B": [
        {
          "name": "Qwen3.6-35B-A3B",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B-FP8": [
        {
          "name": "Qwen3.6-35B-A3B (fp8)",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B-FP8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3.6-35B-A3B-NVFP4": [
        {
          "name": "Qwen3.6-35B-A3B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.6-35B-A3B-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "XiaomiMiMo/MiMo-V2-Flash": [
        {
          "name": "MiMo-V2-Flash",
          "huggingface_id": "XiaomiMiMo/MiMo-V2-Flash",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "baidu/ERNIE-4.5-VL-28B-A3B-PT": [
        {
          "name": "ERNIE-4.5-VL",
          "huggingface_id": "baidu/ERNIE-4.5-VL-28B-A3B-PT",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "baidu/ERNIE-4.5-VL-424B-A47B-PT": [
        {
          "name": "ERNIE-4.5-VL (424b)",
          "huggingface_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "424b"
        },
        {
          "name": "ERNIE-4.5-VL (424b_fp8)",
          "huggingface_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--quantization": "fp8",
                "--cpu-offload-gb": "50",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--quantization": "fp8",
                "--cpu-offload-gb": "50",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "424b_fp8"
        }
      ],
      "deepseek-ai/DeepSeek-V3": [
        {
          "name": "DeepSeek-V3",
          "huggingface_id": "deepseek-ai/DeepSeek-V3",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default",
          "extra_envs": {
            "VLLM_RPC_TIMEOUT": "1800000"
          }
        }
      ],
      "nvidia/DeepSeek-V3-FP4": [
        {
          "name": "DeepSeek-V3 (fp4)",
          "huggingface_id": "nvidia/DeepSeek-V3-FP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Flash": [
        {
          "name": "DeepSeek-V4-Flash",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Flash",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Flash-NVFP4": [
        {
          "name": "DeepSeek-V4-Flash (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Flash-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro": [
        {
          "name": "DeepSeek-V4-Pro",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Pro-NVFP4": [
        {
          "name": "DeepSeek-V4-Pro (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Pro-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro-DSpark": [
        {
          "name": "DeepSeek-V4-Pro (dspark)",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro-DSpark",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "dspark"
        }
      ],
      "inclusionAI/Ring-1T-FP8": [
        {
          "name": "Ring-1T-FP8",
          "huggingface_id": "inclusionAI/Ring-1T-FP8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S1": [
        {
          "name": "Intern-S1",
          "huggingface_id": "internlm/Intern-S1",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S1-FP8": [
        {
          "name": "Intern-S1 (fp8)",
          "huggingface_id": "internlm/Intern-S1-FP8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Ministral-3-8B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (8b)",
          "huggingface_id": "mistralai/Ministral-3-8B-Instruct-2512",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "8b"
        }
      ],
      "mistralai/Ministral-3-3B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (3b)",
          "huggingface_id": "mistralai/Ministral-3-3B-Instruct-2512",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "3b"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512-FP8": [
        {
          "name": "Ministral-3-Instruct (fp8)",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512-FP8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "moonshotai/Kimi-K2-Instruct": [
        {
          "name": "Kimi-K2-Instruct",
          "huggingface_id": "moonshotai/Kimi-K2-Instruct",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2"
              },
              "extra_envs": {
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-K2.5": [
        {
          "name": "Kimi-K2.5",
          "huggingface_id": "moonshotai/Kimi-K2.5",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.5-NVFP4": [
        {
          "name": "Kimi-K2.5 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.5-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.6": [
        {
          "name": "Kimi-K2.6",
          "huggingface_id": "moonshotai/Kimi-K2.6",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.6-NVFP4": [
        {
          "name": "Kimi-K2.6 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.6-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.7-Code": [
        {
          "name": "Kimi-K2.7-Code",
          "huggingface_id": "moonshotai/Kimi-K2.7-Code",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "openai/gpt-oss-120b": [
        {
          "name": "GPT-OSS",
          "huggingface_id": "openai/gpt-oss-120b",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "amd/gpt-oss-120b-w-mxfp4-a-fp8": [
        {
          "name": "GPT-OSS (amd_fp8)",
          "huggingface_id": "amd/gpt-oss-120b-w-mxfp4-a-fp8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "amd_fp8"
        }
      ],
      "openai/gpt-oss-20b": [
        {
          "name": "GPT-OSS 20B",
          "huggingface_id": "openai/gpt-oss-20b",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hunyuan-A13B-Instruct": [
        {
          "name": "Hunyuan-A13B-Instruct",
          "huggingface_id": "tencent/Hunyuan-A13B-Instruct",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hunyuan-A13B-Instruct-FP8": [
        {
          "name": "Hunyuan-A13B-Instruct (fp8)",
          "huggingface_id": "tencent/Hunyuan-A13B-Instruct-FP8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "tencent/Hy3-preview": [
        {
          "name": "Hy3-preview",
          "huggingface_id": "tencent/Hy3-preview",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3": [
        {
          "name": "Hy3",
          "huggingface_id": "tencent/Hy3",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3-FP8": [
        {
          "name": "Hy3 (fp8)",
          "huggingface_id": "tencent/Hy3-FP8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "thinkingmachines/Inkling-NVFP4": [
        {
          "name": "TML Inkling",
          "huggingface_id": "thinkingmachines/Inkling-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "thinkingmachines/Inkling": [
        {
          "name": "TML Inkling (bf16)",
          "huggingface_id": "thinkingmachines/Inkling",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "zai-org/GLM-4.5": [
        {
          "name": "GLM-4.5",
          "huggingface_id": "zai-org/GLM-4.5",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.5-FP8": [
        {
          "name": "GLM-4.5 (fp8)",
          "huggingface_id": "zai-org/GLM-4.5-FP8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.6": [
        {
          "name": "GLM-4.6",
          "huggingface_id": "zai-org/GLM-4.6",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.6-FP8": [
        {
          "name": "GLM-4.6 (fp8)",
          "huggingface_id": "zai-org/GLM-4.6-FP8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.7": [
        {
          "name": "GLM-4.7",
          "huggingface_id": "zai-org/GLM-4.7",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.7-FP8": [
        {
          "name": "GLM-4.7 (fp8)",
          "huggingface_id": "zai-org/GLM-4.7-FP8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-4.7-NVFP4": [
        {
          "name": "GLM-4.7 (nvfp4)",
          "huggingface_id": "nvidia/GLM-4.7-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.2-FP8": [
        {
          "name": "GLM-5.2",
          "huggingface_id": "zai-org/GLM-5.2-FP8",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/GLM-5.2-NVFP4": [
        {
          "name": "GLM-5.2 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5.2-NVFP4",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.2": [
        {
          "name": "GLM-5.2 (bf16)",
          "huggingface_id": "zai-org/GLM-5.2",
          "accelerators": "mi355x",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ]
    },
    "trillium": {
      "google/gemma-4-26B-A4B-it": [
        {
          "name": "Gemma 4 26B-A4B IT",
          "huggingface_id": "google/gemma-4-26B-A4B-it",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-NVFP4",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it": [
        {
          "name": "Gemma 4 31B IT",
          "huggingface_id": "google/gemma-4-31B-it",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-31B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 31B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-31B-it-FP8-dynamic",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/gemma-4-31B-it-NVFP4": [
        {
          "name": "Gemma 4 31B IT (nvfp4)",
          "huggingface_id": "nvidia/gemma-4-31B-it-NVFP4",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 31B IT (w4a16)",
          "huggingface_id": "google/gemma-4-31B-it-qat-w4a16-ct",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it": [
        {
          "name": "Gemma 4 E2B IT",
          "huggingface_id": "google/gemma-4-E2B-it",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E2B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E2B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E2B-it-FP8-dynamic",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E2B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E2B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-w4a16-ct",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E2B IT (mobile)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-mobile-ct",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "google/gemma-4-E4B-it": [
        {
          "name": "Gemma 4 E4B IT",
          "huggingface_id": "google/gemma-4-E4B-it",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E4B-it-FP8-dynamic",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E4B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E4B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-w4a16-ct",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E4B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E4B IT (mobile)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-mobile-ct",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "Qwen/Qwen2.5-VL-7B-Instruct": [
        {
          "name": "Qwen2.5-VL-7B-Instruct",
          "huggingface_id": "Qwen/Qwen2.5-VL-7B-Instruct",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen2.5-VL-7B-Instruct-AWQ": [
        {
          "name": "Qwen2.5-VL-7B-Instruct (awq)",
          "huggingface_id": "Qwen/Qwen2.5-VL-7B-Instruct-AWQ",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "awq",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen3-32B": [
        {
          "name": "Qwen3-32B",
          "huggingface_id": "Qwen/Qwen3-32B",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-32B-FP8": [
        {
          "name": "Qwen3-32B (fp8)",
          "huggingface_id": "Qwen/Qwen3-32B-FP8",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3-32B-AWQ": [
        {
          "name": "Qwen3-32B (awq)",
          "huggingface_id": "Qwen/Qwen3-32B-AWQ",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "awq",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen3-4B": [
        {
          "name": "Qwen3-4B",
          "huggingface_id": "Qwen/Qwen3-4B",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-4B-FP8": [
        {
          "name": "Qwen3-4B (fp8)",
          "huggingface_id": "Qwen/Qwen3-4B-FP8",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "meta-llama/Llama-3.3-70B-Instruct": [
        {
          "name": "Llama-3.3-70B",
          "huggingface_id": "meta-llama/Llama-3.3-70B-Instruct",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-FP8": [
        {
          "name": "Llama-3.3-70B (fp8)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-FP8",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-NVFP4": [
        {
          "name": "Llama-3.3-70B (nvfp4)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-NVFP4",
          "accelerators": "trillium",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ]
    },
    "ironwood": {
      "google/gemma-4-26B-A4B-it": [
        {
          "name": "Gemma 4 26B-A4B IT",
          "huggingface_id": "google/gemma-4-26B-A4B-it",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-NVFP4",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it": [
        {
          "name": "Gemma 4 31B IT",
          "huggingface_id": "google/gemma-4-31B-it",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-31B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 31B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-31B-it-FP8-dynamic",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/gemma-4-31B-it-NVFP4": [
        {
          "name": "Gemma 4 31B IT (nvfp4)",
          "huggingface_id": "nvidia/gemma-4-31B-it-NVFP4",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 31B IT (w4a16)",
          "huggingface_id": "google/gemma-4-31B-it-qat-w4a16-ct",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it": [
        {
          "name": "Gemma 4 E2B IT",
          "huggingface_id": "google/gemma-4-E2B-it",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E2B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E2B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E2B-it-FP8-dynamic",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E2B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E2B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-w4a16-ct",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E2B IT (mobile)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-mobile-ct",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "google/gemma-4-E4B-it": [
        {
          "name": "Gemma 4 E4B IT",
          "huggingface_id": "google/gemma-4-E4B-it",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E4B-it-FP8-dynamic",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E4B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E4B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-w4a16-ct",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E4B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E4B IT (mobile)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-mobile-ct",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "Qwen/Qwen3-32B": [
        {
          "name": "Qwen3-32B",
          "huggingface_id": "Qwen/Qwen3-32B",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-32B-FP8": [
        {
          "name": "Qwen3-32B (fp8)",
          "huggingface_id": "Qwen/Qwen3-32B-FP8",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3-32B-AWQ": [
        {
          "name": "Qwen3-32B (awq)",
          "huggingface_id": "Qwen/Qwen3-32B-AWQ",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "awq",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "openai/gpt-oss-120b": [
        {
          "name": "GPT-OSS",
          "huggingface_id": "openai/gpt-oss-120b",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "amd/gpt-oss-120b-w-mxfp4-a-fp8": [
        {
          "name": "GPT-OSS (amd_fp8)",
          "huggingface_id": "amd/gpt-oss-120b-w-mxfp4-a-fp8",
          "accelerators": "ironwood",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "amd_fp8"
        }
      ]
    },
    "xeon6": {
      "google/gemma-4-26B-A4B-it": [
        {
          "name": "Gemma 4 26B-A4B IT",
          "huggingface_id": "google/gemma-4-26B-A4B-it",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_CPU_KVCACHE_SPACE": "40",
                "VLLM_CPU_ATTN_SPLIT_KV": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_CPU_KVCACHE_SPACE": "40",
                "VLLM_CPU_ATTN_SPLIT_KV": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_CPU_KVCACHE_SPACE": "40",
                "VLLM_CPU_ATTN_SPLIT_KV": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_CPU_KVCACHE_SPACE": "40",
                "VLLM_CPU_ATTN_SPLIT_KV": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-NVFP4",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_CPU_KVCACHE_SPACE": "40",
                "VLLM_CPU_ATTN_SPLIT_KV": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_CPU_KVCACHE_SPACE": "40",
                "VLLM_CPU_ATTN_SPLIT_KV": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-E2B-it": [
        {
          "name": "Gemma 4 E2B IT",
          "huggingface_id": "google/gemma-4-E2B-it",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_CPU_KVCACHE_SPACE": "40",
                "VLLM_CPU_ATTN_SPLIT_KV": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E2B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E2B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E2B-it-FP8-dynamic",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_CPU_KVCACHE_SPACE": "40",
                "VLLM_CPU_ATTN_SPLIT_KV": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E2B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E2B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-w4a16-ct",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_CPU_KVCACHE_SPACE": "40",
                "VLLM_CPU_ATTN_SPLIT_KV": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E2B IT (mobile)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-mobile-ct",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_CPU_KVCACHE_SPACE": "40",
                "VLLM_CPU_ATTN_SPLIT_KV": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "google/gemma-4-E4B-it": [
        {
          "name": "Gemma 4 E4B IT",
          "huggingface_id": "google/gemma-4-E4B-it",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_CPU_KVCACHE_SPACE": "40",
                "VLLM_CPU_ATTN_SPLIT_KV": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E4B-it-FP8-dynamic",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_CPU_KVCACHE_SPACE": "40",
                "VLLM_CPU_ATTN_SPLIT_KV": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E4B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E4B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-w4a16-ct",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_CPU_KVCACHE_SPACE": "40",
                "VLLM_CPU_ATTN_SPLIT_KV": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E4B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E4B IT (mobile)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-mobile-ct",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_CPU_KVCACHE_SPACE": "40",
                "VLLM_CPU_ATTN_SPLIT_KV": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "Qwen/Qwen3-4B": [
        {
          "name": "Qwen3-4B",
          "huggingface_id": "Qwen/Qwen3-4B",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-4B-FP8": [
        {
          "name": "Qwen3-4B (fp8)",
          "huggingface_id": "Qwen/Qwen3-4B-FP8",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3.5-35B-A3B": [
        {
          "name": "Qwen3.5-35B-A3B",
          "huggingface_id": "Qwen/Qwen3.5-35B-A3B",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-35B-A3B-FP8": [
        {
          "name": "Qwen3.5-35B-A3B (fp8)",
          "huggingface_id": "Qwen/Qwen3.5-35B-A3B-FP8",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3.5-35B-A3B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-35B-A3B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-35B-A3B-GPTQ-Int4",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-4B": [
        {
          "name": "Qwen3.5-4B",
          "huggingface_id": "Qwen/Qwen3.5-4B",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "meta-llama/Llama-3.3-70B-Instruct": [
        {
          "name": "Llama-3.3-70B",
          "huggingface_id": "meta-llama/Llama-3.3-70B-Instruct",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-FP8": [
        {
          "name": "Llama-3.3-70B (fp8)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-FP8",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-NVFP4": [
        {
          "name": "Llama-3.3-70B (nvfp4)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-NVFP4",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "meta-llama/Llama-4-Scout-17B-16E-Instruct": [
        {
          "name": "Llama-4-Scout",
          "huggingface_id": "meta-llama/Llama-4-Scout-17B-16E-Instruct",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8": [
        {
          "name": "Llama-4-Scout (fp8)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4": [
        {
          "name": "Llama-4-Scout (nvfp4)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4",
          "accelerators": "xeon6",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ]
    },
    "xeon5": {},
    "rtxpro-6000-blackwell-server-edition": {
      "nvidia/Gemma-4-26B-A4B-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-26B-A4B-NVFP4",
          "accelerators": "rtxpro-6000-blackwell-server-edition",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--gpu-memory-utilization": "0.8"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": "",
                "--gpu-memory-utilization": "0.8"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-31B-IT-NVFP4": [
        {
          "name": "Gemma 4 31B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-31B-IT-NVFP4",
          "accelerators": "rtxpro-6000-blackwell-server-edition",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}",
                "--gpu-memory-utilization": "0.8"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ]
    },
    "amd": {
      "ByteDance-Seed/Seed-OSS-36B-Instruct": [
        {
          "name": "Seed-OSS-36B",
          "huggingface_id": "ByteDance-Seed/Seed-OSS-36B-Instruct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "seed_oss"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "google/diffusiongemma-26B-A4B-it": [
        {
          "name": "DiffusionGemma 26B-A4B IT",
          "huggingface_id": "google/diffusiongemma-26B-A4B-it",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/diffusiongemma-26B-A4B-it-FP8-dynamic": [
        {
          "name": "DiffusionGemma 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/diffusiongemma-26B-A4B-it-FP8-dynamic",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/diffusiongemma-26B-A4B-it-NVFP4": [
        {
          "name": "DiffusionGemma 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/diffusiongemma-26B-A4B-it-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-12B-it": [
        {
          "name": "Gemma 4 12B IT",
          "huggingface_id": "google/gemma-4-12B-it",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-12B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "google/gemma-4-12B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 12B IT (w4a16)",
          "huggingface_id": "google/gemma-4-12B-it-qat-w4a16-ct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-12B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-26B-A4B-it": [
        {
          "name": "Gemma 4 26B-A4B IT",
          "huggingface_id": "google/gemma-4-26B-A4B-it",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it": [
        {
          "name": "Gemma 4 31B IT",
          "huggingface_id": "google/gemma-4-31B-it",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-31B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 31B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-31B-it-FP8-dynamic",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/gemma-4-31B-it-NVFP4": [
        {
          "name": "Gemma 4 31B IT (nvfp4)",
          "huggingface_id": "nvidia/gemma-4-31B-it-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 31B IT (w4a16)",
          "huggingface_id": "google/gemma-4-31B-it-qat-w4a16-ct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it": [
        {
          "name": "Gemma 4 E2B IT",
          "huggingface_id": "google/gemma-4-E2B-it",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E2B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E2B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E2B-it-FP8-dynamic",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E2B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E2B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-w4a16-ct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E2B IT (mobile)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-mobile-ct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "google/gemma-4-E4B-it": [
        {
          "name": "Gemma 4 E4B IT",
          "huggingface_id": "google/gemma-4-E4B-it",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E4B-it-FP8-dynamic",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E4B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E4B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-w4a16-ct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E4B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E4B IT (mobile)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-mobile-ct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "google/translategemma-27b-it": [
        {
          "name": "TranslateGemma 27B IT",
          "huggingface_id": "google/translategemma-27b-it",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Infomaniak-AI/vllm-translategemma-27b-it": [
        {
          "name": "TranslateGemma 27B IT (vllm_optimized)",
          "huggingface_id": "Infomaniak-AI/vllm-translategemma-27b-it",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "vllm_optimized"
        }
      ],
      "google/translategemma-4b-it": [
        {
          "name": "TranslateGemma 27B IT (small_4b)",
          "huggingface_id": "google/translategemma-4b-it",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "small_4b"
        }
      ],
      "JetBrains/Mellum2-12B-A2.5B-Instruct": [
        {
          "name": "Mellum2-12B-A2.5B-Instruct",
          "huggingface_id": "JetBrains/Mellum2-12B-A2.5B-Instruct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "JetBrains/Mellum2-12B-A2.5B-Thinking": [
        {
          "name": "Mellum2-12B-A2.5B-Thinking",
          "huggingface_id": "JetBrains/Mellum2-12B-A2.5B-Thinking",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Base": [
        {
          "name": "LFM2.5 1.2B Base",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Base",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Instruct": [
        {
          "name": "LFM2.5 1.2B Instruct",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Instruct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-JP-202606": [
        {
          "name": "LFM2.5 1.2B JP (202606)",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-JP-202606",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-JP": [
        {
          "name": "LFM2.5 1.2B JP",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-JP",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Thinking": [
        {
          "name": "LFM2.5 1.2B Thinking",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Thinking",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-230M": [
        {
          "name": "LFM2.5 230M",
          "huggingface_id": "LiquidAI/LFM2.5-230M",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-350M": [
        {
          "name": "LFM2.5 350M",
          "huggingface_id": "LiquidAI/LFM2.5-350M",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-8B-A1B": [
        {
          "name": "LFM2.5 8B A1B",
          "huggingface_id": "LiquidAI/LFM2.5-8B-A1B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-VL-1.6B": [
        {
          "name": "LFM2.5 VL 1.6B",
          "huggingface_id": "LiquidAI/LFM2.5-VL-1.6B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-VL-450M": [
        {
          "name": "LFM2.5 VL 450M",
          "huggingface_id": "LiquidAI/LFM2.5-VL-450M",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M2.1": [
        {
          "name": "MiniMax-M2.1",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.1",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M2.5": [
        {
          "name": "MiniMax-M2.5",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.5",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.5-NVFP4": [
        {
          "name": "MiniMax-M2.5 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.5-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2.7": [
        {
          "name": "MiniMax-M2.7",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.7",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.7-NVFP4": [
        {
          "name": "MiniMax-M2.7 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.7-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2": [
        {
          "name": "MiniMax-M2",
          "huggingface_id": "MiniMaxAI/MiniMax-M2",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/MiniMax-M2-NVFP4": [
        {
          "name": "MiniMax-M2 (nvfp4)",
          "huggingface_id": "RedHatAI/MiniMax-M2-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--attention-backend": "ROCM_AITER_FA",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M3": [
        {
          "name": "MiniMax-M3",
          "huggingface_id": "MiniMaxAI/MiniMax-M3",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M3-MXFP8": [
        {
          "name": "MiniMax-M3 (mxfp8)",
          "huggingface_id": "MiniMaxAI/MiniMax-M3-MXFP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp8"
        }
      ],
      "nvidia/MiniMax-M3-NVFP4": [
        {
          "name": "MiniMax-M3 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M3-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0",
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0",
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "amd/MiniMax-M3-MXFP4": [
        {
          "name": "MiniMax-M3 (mxfp4)",
          "huggingface_id": "amd/MiniMax-M3-MXFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--attention-backend": "TRITON_ATTN",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_BREAKABLE_CUDAGRAPH": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp4"
        }
      ],
      "OpenGVLab/InternVL3_5-8B": [
        {
          "name": "InternVL3.5",
          "huggingface_id": "OpenGVLab/InternVL3_5-8B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-SoundEffect": [
        {
          "name": "MOSS-SoundEffect",
          "huggingface_id": "OpenMOSS-Team/MOSS-SoundEffect",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_sound_effect.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-TTS-Realtime": [
        {
          "name": "MOSS-TTS-Realtime",
          "huggingface_id": "OpenMOSS-Team/MOSS-TTS-Realtime",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-TTS": [
        {
          "name": "MOSS-TTS",
          "huggingface_id": "OpenMOSS-Team/MOSS-TTS",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_tts.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-TTSD-v1.0": [
        {
          "name": "MOSS-TTSD-v1.0",
          "huggingface_id": "OpenMOSS-Team/MOSS-TTSD-v1.0",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_ttsd.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-Transcribe-Diarize": [
        {
          "name": "MOSS-Transcribe-Diarize",
          "huggingface_id": "OpenMOSS-Team/MOSS-Transcribe-Diarize",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-VoiceGenerator": [
        {
          "name": "MOSS-VoiceGenerator",
          "huggingface_id": "OpenMOSS-Team/MOSS-VoiceGenerator",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_voice_generator.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "PaddlePaddle/PaddleOCR-VL-1.5": [
        {
          "name": "PaddleOCR-VL-1.5",
          "huggingface_id": "PaddlePaddle/PaddleOCR-VL-1.5",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-batched-tokens": "16384",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "PaddlePaddle/PaddleOCR-VL-1.6": [
        {
          "name": "PaddleOCR-VL-1.6",
          "huggingface_id": "PaddlePaddle/PaddleOCR-VL-1.6",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-batched-tokens": "16384",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "PaddlePaddle/PaddleOCR-VL": [
        {
          "name": "PaddleOCR-VL",
          "huggingface_id": "PaddlePaddle/PaddleOCR-VL",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-batched-tokens": "16384",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen-Image": [
        {
          "name": "Qwen-Image",
          "huggingface_id": "Qwen/Qwen-Image",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        },
        {
          "name": "Qwen-Image (fp8)",
          "huggingface_id": "Qwen/Qwen-Image",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "fp8",
                "--ignored-layers": "img_mlp"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        },
        {
          "name": "Qwen-Image (int8)",
          "huggingface_id": "Qwen/Qwen-Image",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "int8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int8"
        }
      ],
      "Qwen/Qwen2.5-VL-72B-Instruct": [
        {
          "name": "Qwen2.5-VL-72B-Instruct",
          "huggingface_id": "Qwen/Qwen2.5-VL-72B-Instruct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--limit-mm-per-prompt": "{\"image\":2,\"video\":0}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen2.5-VL-72B-Instruct-AWQ": [
        {
          "name": "Qwen2.5-VL-72B-Instruct (awq)",
          "huggingface_id": "Qwen/Qwen2.5-VL-72B-Instruct-AWQ",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--limit-mm-per-prompt": "{\"image\":2,\"video\":0}",
                "--quantization": "awq"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen2.5-VL-7B-Instruct": [
        {
          "name": "Qwen2.5-VL-7B-Instruct",
          "huggingface_id": "Qwen/Qwen2.5-VL-7B-Instruct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen2.5-VL-7B-Instruct-AWQ": [
        {
          "name": "Qwen2.5-VL-7B-Instruct (awq)",
          "huggingface_id": "Qwen/Qwen2.5-VL-7B-Instruct-AWQ",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "awq",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen3-235B-A22B-Instruct-2507": [
        {
          "name": "Qwen3-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-Instruct-2507",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-235B-A22B-FP8": [
        {
          "name": "Qwen3-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4": [
        {
          "name": "Qwen3-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-32B": [
        {
          "name": "Qwen3-32B",
          "huggingface_id": "Qwen/Qwen3-32B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-32B-FP8": [
        {
          "name": "Qwen3-32B (fp8)",
          "huggingface_id": "Qwen/Qwen3-32B-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3-32B-AWQ": [
        {
          "name": "Qwen3-32B (awq)",
          "huggingface_id": "Qwen/Qwen3-32B-AWQ",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "awq",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen3-4B": [
        {
          "name": "Qwen3-4B",
          "huggingface_id": "Qwen/Qwen3-4B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-4B-FP8": [
        {
          "name": "Qwen3-4B (fp8)",
          "huggingface_id": "Qwen/Qwen3-4B-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3-ASR-1.7B": [
        {
          "name": "Qwen3-ASR-1.7B",
          "huggingface_id": "Qwen/Qwen3-ASR-1.7B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (voicedesign-1.7b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "voicedesign-1.7b"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-1.7B-Base": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (base-1.7b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-1.7B-Base",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base-1.7b"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (customvoice-0.6b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "customvoice-0.6b"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-0.6B-Base": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (base-0.6b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-0.6B-Base",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base-0.6b"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--async-scheduling": "",
                "--enable-chunked-prefill": "",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_WORKER_MULTIPROC_METHOD": "spawn",
                "HIP_FORCE_DEV_KERNARG": "1",
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_SHUFFLE_KV_CACHE_LAYOUT": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-0.8B": [
        {
          "name": "Qwen3.5-0.8B",
          "huggingface_id": "Qwen/Qwen3.5-0.8B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-122B-A10B": [
        {
          "name": "Qwen3.5-122B-A10B",
          "huggingface_id": "Qwen/Qwen3.5-122B-A10B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-122B-A10B-FP8": [
        {
          "name": "Qwen3.5-122B-A10B (fp8)",
          "huggingface_id": "Qwen/Qwen3.5-122B-A10B-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3.5-122B-A10B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-122B-A10B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-122B-A10B-GPTQ-Int4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-27B": [
        {
          "name": "Qwen3.5-27B",
          "huggingface_id": "Qwen/Qwen3.5-27B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-27B-FP8": [
        {
          "name": "Qwen3.5-27B (fp8)",
          "huggingface_id": "Qwen/Qwen3.5-27B-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3.5-27B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-27B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-27B-GPTQ-Int4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-2B": [
        {
          "name": "Qwen3.5-2B",
          "huggingface_id": "Qwen/Qwen3.5-2B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-35B-A3B": [
        {
          "name": "Qwen3.5-35B-A3B",
          "huggingface_id": "Qwen/Qwen3.5-35B-A3B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-35B-A3B-FP8": [
        {
          "name": "Qwen3.5-35B-A3B (fp8)",
          "huggingface_id": "Qwen/Qwen3.5-35B-A3B-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3.5-35B-A3B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-35B-A3B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-35B-A3B-GPTQ-Int4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B": [
        {
          "name": "Qwen3.5-397B",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Qwen3.5-397B-A17B-NVFP4": [
        {
          "name": "Qwen3.5-397B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.5-397B-A17B-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-397B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-4B": [
        {
          "name": "Qwen3.5-4B",
          "huggingface_id": "Qwen/Qwen3.5-4B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-9B": [
        {
          "name": "Qwen3.5-9B",
          "huggingface_id": "Qwen/Qwen3.5-9B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-27B": [
        {
          "name": "Qwen3.6-27B",
          "huggingface_id": "Qwen/Qwen3.6-27B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-27B-FP8": [
        {
          "name": "Qwen3.6-27B (fp8)",
          "huggingface_id": "Qwen/Qwen3.6-27B-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "512",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3.6-27B-NVFP4": [
        {
          "name": "Qwen3.6-27B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.6-27B-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B": [
        {
          "name": "Qwen3.6-35B-A3B",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B-FP8": [
        {
          "name": "Qwen3.6-35B-A3B (fp8)",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3.6-35B-A3B-NVFP4": [
        {
          "name": "Qwen3.6-35B-A3B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.6-35B-A3B-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3Guard-Gen-8B": [
        {
          "name": "Qwen3Guard-Gen-8B",
          "huggingface_id": "Qwen/Qwen3Guard-Gen-8B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3Guard-Gen-4B": [
        {
          "name": "Qwen3Guard-Gen-8B (small_4b)",
          "huggingface_id": "Qwen/Qwen3Guard-Gen-4B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "small_4b"
        }
      ],
      "Qwen/Qwen3Guard-Gen-0.6B": [
        {
          "name": "Qwen3Guard-Gen-8B (tiny_0_6b)",
          "huggingface_id": "Qwen/Qwen3Guard-Gen-0.6B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "tiny_0_6b"
        }
      ],
      "Wan-AI/Wan2.2-T2V-A14B-Diffusers": [
        {
          "name": "Wan2.2",
          "huggingface_id": "Wan-AI/Wan2.2-T2V-A14B-Diffusers",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "XiaomiMiMo/MiMo-V2-Flash": [
        {
          "name": "MiMo-V2-Flash",
          "huggingface_id": "XiaomiMiMo/MiMo-V2-Flash",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "XiaomiMiMo/MiMo-V2.5-Pro": [
        {
          "name": "MiMo-V2.5-Pro",
          "huggingface_id": "XiaomiMiMo/MiMo-V2.5-Pro",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "XiaomiMiMo/MiMo-V2.5": [
        {
          "name": "MiMo-V2.5",
          "huggingface_id": "XiaomiMiMo/MiMo-V2.5",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "arcee-ai/Trinity-Large-Thinking": [
        {
          "name": "Trinity-Large-Thinking",
          "huggingface_id": "arcee-ai/Trinity-Large-Thinking",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "arcee-ai/Trinity-Large-Thinking-NVFP4": [
        {
          "name": "Trinity-Large-Thinking (nvfp4)",
          "huggingface_id": "arcee-ai/Trinity-Large-Thinking-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "baidu/ERNIE-4.5-21B-A3B-PT": [
        {
          "name": "ERNIE-4.5",
          "huggingface_id": "baidu/ERNIE-4.5-21B-A3B-PT",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--gpu-memory-utilization": "0.9",
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--gpu-memory-utilization": "0.9",
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "baidu/ERNIE-4.5-300B-A47B-PT": [
        {
          "name": "ERNIE-4.5 (300b)",
          "huggingface_id": "baidu/ERNIE-4.5-300B-A47B-PT",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--gpu-memory-utilization": "0.95",
                "--quantization": "fp8",
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--gpu-memory-utilization": "0.95",
                "--quantization": "fp8",
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "300b"
        }
      ],
      "baidu/ERNIE-4.5-VL-28B-A3B-PT": [
        {
          "name": "ERNIE-4.5-VL",
          "huggingface_id": "baidu/ERNIE-4.5-VL-28B-A3B-PT",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "baidu/ERNIE-4.5-VL-424B-A47B-PT": [
        {
          "name": "ERNIE-4.5-VL (424b)",
          "huggingface_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "424b"
        },
        {
          "name": "ERNIE-4.5-VL (424b_fp8)",
          "huggingface_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--quantization": "fp8",
                "--cpu-offload-gb": "50",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--quantization": "fp8",
                "--cpu-offload-gb": "50",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "424b_fp8"
        }
      ],
      "baidu/Unlimited-OCR": [
        {
          "name": "Unlimited-OCR",
          "huggingface_id": "baidu/Unlimited-OCR",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--logits_processors": "vllm.model_executor.models.unlimited_ocr:NGramPerReqLogitsProcessor",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "bosonai/higgs-audio-v3-tts-4b": [
        {
          "name": "Higgs Audio V3 TTS",
          "huggingface_id": "bosonai/higgs-audio-v3-tts-4b",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-OCR-2": [
        {
          "name": "DeepSeek-OCR-2",
          "huggingface_id": "deepseek-ai/DeepSeek-OCR-2",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--logits_processors": "vllm.model_executor.models.deepseek_ocr:NGramPerReqLogitsProcessor",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-OCR": [
        {
          "name": "DeepSeek-OCR",
          "huggingface_id": "deepseek-ai/DeepSeek-OCR",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--logits_processors": "vllm.model_executor.models.deepseek_ocr:NGramPerReqLogitsProcessor",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-R1": [
        {
          "name": "DeepSeek-R1",
          "huggingface_id": "deepseek-ai/DeepSeek-R1",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default",
          "extra_envs": {
            "VLLM_RPC_TIMEOUT": "1800000"
          }
        }
      ],
      "deepseek-ai/DeepSeek-R1-0528": [
        {
          "name": "DeepSeek-R1 (r1_0528)",
          "huggingface_id": "deepseek-ai/DeepSeek-R1-0528",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "r1_0528"
        }
      ],
      "nvidia/DeepSeek-R1-0528-NVFP4-v2": [
        {
          "name": "DeepSeek-R1 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-R1-0528-NVFP4-v2",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3.1": [
        {
          "name": "DeepSeek-V3.1",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.1",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3.1-NVFP4": [
        {
          "name": "DeepSeek-V3.1 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V3.1-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3.2-Exp": [
        {
          "name": "DeepSeek-V3.2-Exp",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.2-Exp",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--block-size": "1",
                "--kv-cache-dtype": "bfloat16",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "32768",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_RPC_TIMEOUT": "18000000"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--block-size": "1",
                "--kv-cache-dtype": "bfloat16",
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "32768",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_RPC_TIMEOUT": "18000000"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-V3.2": [
        {
          "name": "DeepSeek-V3.2",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.2",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3.2-NVFP4": [
        {
          "name": "DeepSeek-V3.2 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V3.2-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--kv-cache-dtype": "fp8",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--kv-cache-dtype": "fp8",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3": [
        {
          "name": "DeepSeek-V3",
          "huggingface_id": "deepseek-ai/DeepSeek-V3",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default",
          "extra_envs": {
            "VLLM_RPC_TIMEOUT": "1800000"
          }
        }
      ],
      "nvidia/DeepSeek-V3-FP4": [
        {
          "name": "DeepSeek-V3 (fp4)",
          "huggingface_id": "nvidia/DeepSeek-V3-FP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "SAFETENSORS_FAST_GPU": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Flash": [
        {
          "name": "DeepSeek-V4-Flash",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Flash",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Flash-NVFP4": [
        {
          "name": "DeepSeek-V4-Flash (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Flash-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro": [
        {
          "name": "DeepSeek-V4-Pro",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Pro-NVFP4": [
        {
          "name": "DeepSeek-V4-Pro (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Pro-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro-DSpark": [
        {
          "name": "DeepSeek-V4-Pro (dspark)",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro-DSpark",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--distributed-executor-backend": "mp",
                "--gpu-memory-utilization": "0.9",
                "--max-num-seqs": "512",
                "--max-num-batched-tokens": "8192",
                "--compilation-config": "{\"mode\": 3, \"cudagraph_mode\": \"FULL_DECODE_ONLY\"}",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "dspark"
        }
      ],
      "fishaudio/s2-pro": [
        {
          "name": "Fish Speech S2 Pro",
          "huggingface_id": "fishaudio/s2-pro",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ling-2.6-1T": [
        {
          "name": "Ling-2.6-1T",
          "huggingface_id": "inclusionAI/Ling-2.6-1T",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ling-2.6-flash": [
        {
          "name": "Ling-2.6-flash",
          "huggingface_id": "inclusionAI/Ling-2.6-flash",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ming-omni-tts-0.5B": [
        {
          "name": "Ming-omni-tts-0.5B",
          "huggingface_id": "inclusionAI/Ming-omni-tts-0.5B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--enforce-eager": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ring-1T-FP8": [
        {
          "name": "Ring-1T-FP8",
          "huggingface_id": "inclusionAI/Ring-1T-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ring-2.6-1T": [
        {
          "name": "Ring-2.6-1T",
          "huggingface_id": "inclusionAI/Ring-2.6-1T",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S1": [
        {
          "name": "Intern-S1",
          "huggingface_id": "internlm/Intern-S1",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S1-FP8": [
        {
          "name": "Intern-S1 (fp8)",
          "huggingface_id": "internlm/Intern-S1-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "internlm/Intern-S2-Preview": [
        {
          "name": "Intern-S2-Preview",
          "huggingface_id": "internlm/Intern-S2-Preview",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S2-Preview-FP8": [
        {
          "name": "Intern-S2-Preview (fp8)",
          "huggingface_id": "internlm/Intern-S2-Preview-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-retrieval": [
        {
          "name": "Jina Embeddings v5 Text Small",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-retrieval",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-text-matching": [
        {
          "name": "Jina Embeddings v5 Text Small (text_matching)",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-text-matching",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "text_matching"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-classification": [
        {
          "name": "Jina Embeddings v5 Text Small (classification)",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-classification",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "classification"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-clustering": [
        {
          "name": "Jina Embeddings v5 Text Small (clustering)",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-clustering",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "clustering"
        }
      ],
      "jinaai/jina-reranker-m0": [
        {
          "name": "Jina Reranker m0",
          "huggingface_id": "jinaai/jina-reranker-m0",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--gpu-memory-utilization": "0.75",
                "--max-num-seqs": "32"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "meituan-longcat/LongCat-Image-Edit": [
        {
          "name": "LongCat-Image-Edit",
          "huggingface_id": "meituan-longcat/LongCat-Image-Edit",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "meta-llama/Llama-3.3-70B-Instruct": [
        {
          "name": "Llama-3.3-70B",
          "huggingface_id": "meta-llama/Llama-3.3-70B-Instruct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-FP8": [
        {
          "name": "Llama-3.3-70B (fp8)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-NVFP4": [
        {
          "name": "Llama-3.3-70B (nvfp4)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "meta-llama/Llama-4-Scout-17B-16E-Instruct": [
        {
          "name": "Llama-4-Scout",
          "huggingface_id": "meta-llama/Llama-4-Scout-17B-16E-Instruct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "16384",
                "--max-num-seqs": "64"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "16384",
                "--max-num-seqs": "64",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8": [
        {
          "name": "Llama-4-Scout (fp8)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "16384",
                "--max-num-seqs": "64",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "16384",
                "--max-num-seqs": "64",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4": [
        {
          "name": "Llama-4-Scout (nvfp4)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "16384",
                "--max-num-seqs": "64",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--max-num-batched-tokens": "16384",
                "--max-num-seqs": "64",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "microsoft/Phi-4-mini-instruct": [
        {
          "name": "Phi-4",
          "huggingface_id": "microsoft/Phi-4-mini-instruct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "microsoft/Phi-4-mini-reasoning": [
        {
          "name": "Phi-4 (mini_reasoning)",
          "huggingface_id": "microsoft/Phi-4-mini-reasoning",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mini_reasoning"
        }
      ],
      "microsoft/Phi-4-reasoning": [
        {
          "name": "Phi-4 (reasoning)",
          "huggingface_id": "microsoft/Phi-4-reasoning",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "reasoning"
        }
      ],
      "microsoft/Phi-4-multimodal-instruct": [
        {
          "name": "Phi-4 (multimodal)",
          "huggingface_id": "microsoft/Phi-4-multimodal-instruct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "multimodal"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Ministral-3-8B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (8b)",
          "huggingface_id": "mistralai/Ministral-3-8B-Instruct-2512",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "8b"
        }
      ],
      "mistralai/Ministral-3-3B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (3b)",
          "huggingface_id": "mistralai/Ministral-3-3B-Instruct-2512",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "3b"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512-FP8": [
        {
          "name": "Ministral-3-Instruct (fp8)",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "mistralai/Ministral-3-8B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning",
          "huggingface_id": "mistralai/Ministral-3-8B-Reasoning-2512",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Ministral-3-3B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning (3b)",
          "huggingface_id": "mistralai/Ministral-3-3B-Reasoning-2512",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "3b"
        }
      ],
      "mistralai/Ministral-3-14B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning (14b)",
          "huggingface_id": "mistralai/Ministral-3-14B-Reasoning-2512",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "14b"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512": [
        {
          "name": "Mistral-Large-3-675B-Instruct",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4": [
        {
          "name": "Mistral-Large-3-675B-Instruct (nvfp4)",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512-FP8": [
        {
          "name": "Mistral-Large-3-675B-Instruct (fp8)",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--no-enable-prefix-caching": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--no-enable-prefix-caching": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "mistralai/Mistral-Medium-3.5-128B": [
        {
          "name": "Mistral-Medium-3.5",
          "huggingface_id": "mistralai/Mistral-Medium-3.5-128B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Medium-3.5-128B-EAGLE\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Mistral-Small-4-119B-2603": [
        {
          "name": "Mistral-Small-4-119B",
          "huggingface_id": "mistralai/Mistral-Small-4-119B-2603",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "FLASH_ATTN_MLA",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "FLASH_ATTN_MLA",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Mistral-Small-4-119B-2603-NVFP4": [
        {
          "name": "Mistral-Small-4-119B (nvfp4)",
          "huggingface_id": "mistralai/Mistral-Small-4-119B-2603-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "TRITON_MLA",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "TRITON_MLA",
                "--no-enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "mistralai/Voxtral-4B-TTS-2603": [
        {
          "name": "Voxtral-4B-TTS-2603",
          "huggingface_id": "mistralai/Voxtral-4B-TTS-2603",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Voxtral-Mini-4B-Realtime-2602": [
        {
          "name": "Voxtral-Mini-4B-Realtime-2602",
          "huggingface_id": "mistralai/Voxtral-Mini-4B-Realtime-2602",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer-mode": "mistral",
                "--compilation_config": "{\"cudagraph_mode\": \"PIECEWISE\"}"
              },
              "extra_envs": {
                "VLLM_DISABLE_COMPILE_CACHE": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-K2-Instruct": [
        {
          "name": "Kimi-K2-Instruct",
          "huggingface_id": "moonshotai/Kimi-K2-Instruct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2"
              },
              "extra_envs": {
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-K2-Thinking": [
        {
          "name": "Kimi-K2-Thinking",
          "huggingface_id": "moonshotai/Kimi-K2-Thinking",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2-Thinking-NVFP4": [
        {
          "name": "Kimi-K2-Thinking (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2-Thinking-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.5": [
        {
          "name": "Kimi-K2.5",
          "huggingface_id": "moonshotai/Kimi-K2.5",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.5-NVFP4": [
        {
          "name": "Kimi-K2.5 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.5-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.6": [
        {
          "name": "Kimi-K2.6",
          "huggingface_id": "moonshotai/Kimi-K2.6",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.6-NVFP4": [
        {
          "name": "Kimi-K2.6 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.6-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.7-Code": [
        {
          "name": "Kimi-K2.7-Code",
          "huggingface_id": "moonshotai/Kimi-K2.7-Code",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-Linear-48B-A3B-Instruct": [
        {
          "name": "Kimi-Linear-48B-A3B-Instruct",
          "huggingface_id": "moonshotai/Kimi-Linear-48B-A3B-Instruct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Nano": [
        {
          "name": "Cosmos3-Nano",
          "huggingface_id": "nvidia/Cosmos3-Nano",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super-Image2Video": [
        {
          "name": "Cosmos3-Super-Image2Video",
          "huggingface_id": "nvidia/Cosmos3-Super-Image2Video",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super-Text2Image": [
        {
          "name": "Cosmos3-Super-Text2Image",
          "huggingface_id": "nvidia/Cosmos3-Super-Text2Image",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super": [
        {
          "name": "Cosmos3-Super",
          "huggingface_id": "nvidia/Cosmos3-Super",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Nano-30B-A3B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Nano-30B-A3B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_FLASHINFER_MOE_BACKEND": "throughput"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_FLASHINFER_MOE_BACKEND": "throughput"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Nano-4B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Nano-4B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (base_bf16)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base_bf16"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Ultra-550B-A55B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}"
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4": [
        {
          "name": "NVIDIA Nemotron-3-Ultra-550B-A55B (fp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}"
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-FP8": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-NVFP4-QAD": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-NVFP4-QAD",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-FP8": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-NVFP4": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Base": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (base)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Base",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Japanese": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (japanese)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Japanese",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "japanese"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning (fp8)",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning (nvfp4)",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "openai/gpt-oss-120b": [
        {
          "name": "GPT-OSS",
          "huggingface_id": "openai/gpt-oss-120b",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "amd/gpt-oss-120b-w-mxfp4-a-fp8": [
        {
          "name": "GPT-OSS (amd_fp8)",
          "huggingface_id": "amd/gpt-oss-120b-w-mxfp4-a-fp8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "amd_fp8"
        }
      ],
      "openai/gpt-oss-20b": [
        {
          "name": "GPT-OSS 20B",
          "huggingface_id": "openai/gpt-oss-20b",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "ROCM_AITER_UNIFIED_ATTN",
                "-cc.pass_config.fuse_rope_kvcache=True": "",
                "-cc.use_inductor_graph_partition=True": "",
                "--gpu-memory-utilization": "0.95",
                "--block-size=64": "",
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
                "HSA_NO_SCRATCH_RECLAIM": "1",
                "AMDGCN_USE_BUFFER_OPS": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "openbmb/MiniCPM-V-4.6": [
        {
          "name": "MiniCPM-V 4.6",
          "huggingface_id": "openbmb/MiniCPM-V-4.6",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "openbmb/MiniCPM5-1B": [
        {
          "name": "MiniCPM5-1B",
          "huggingface_id": "openbmb/MiniCPM5-1B",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "openbmb/VoxCPM2": [
        {
          "name": "VoxCPM2",
          "huggingface_id": "openbmb/VoxCPM2",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "pfnet/plamo-2-translate": [
        {
          "name": "PLaMo 2 Translate",
          "huggingface_id": "pfnet/plamo-2-translate",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "pfnet/plamo-3-nict-31b-base": [
        {
          "name": "PLaMo 3 NICT 31B Base",
          "huggingface_id": "pfnet/plamo-3-nict-31b-base",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-M.1": [
        {
          "name": "Laguna M.1",
          "huggingface_id": "poolside/Laguna-M.1",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-M.1-FP8": [
        {
          "name": "Laguna M.1 (fp8)",
          "huggingface_id": "poolside/Laguna-M.1-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "poolside/Laguna-M.1-NVFP4": [
        {
          "name": "Laguna M.1 (nvfp4)",
          "huggingface_id": "poolside/Laguna-M.1-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "poolside/Laguna-XS-2.1": [
        {
          "name": "Laguna XS-2.1",
          "huggingface_id": "poolside/Laguna-XS-2.1",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-XS-2.1-FP8": [
        {
          "name": "Laguna XS-2.1 (fp8)",
          "huggingface_id": "poolside/Laguna-XS-2.1-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "poolside/Laguna-XS-2.1-NVFP4": [
        {
          "name": "Laguna XS-2.1 (nvfp4)",
          "huggingface_id": "poolside/Laguna-XS-2.1-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "poolside/Laguna-XS-2.1-INT4": [
        {
          "name": "Laguna XS-2.1 (int4)",
          "huggingface_id": "poolside/Laguna-XS-2.1-INT4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "poolside/Laguna-XS.2": [
        {
          "name": "Laguna XS.2",
          "huggingface_id": "poolside/Laguna-XS.2",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-XS.2-FP8": [
        {
          "name": "Laguna XS.2 (fp8)",
          "huggingface_id": "poolside/Laguna-XS.2-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "poolside/Laguna-XS.2-NVFP4": [
        {
          "name": "Laguna XS.2 (nvfp4)",
          "huggingface_id": "poolside/Laguna-XS.2-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "poolside/Laguna-XS.2-INT4": [
        {
          "name": "Laguna XS.2 (int4)",
          "huggingface_id": "poolside/Laguna-XS.2-INT4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "stabilityai/stable-audio-open-1.0": [
        {
          "name": "Stable Audio Open",
          "huggingface_id": "stabilityai/stable-audio-open-1.0",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enforce-eager": "",
                "--gpu-memory-utilization": "0.9"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stabilityai/stable-diffusion-3.5-medium": [
        {
          "name": "Stable Diffusion 3.5",
          "huggingface_id": "stabilityai/stable-diffusion-3.5-medium",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stabilityai/stable-diffusion-3.5-large": [
        {
          "name": "Stable Diffusion 3.5 (large)",
          "huggingface_id": "stabilityai/stable-diffusion-3.5-large",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "large"
        }
      ],
      "stabilityai/stable-diffusion-3.5-large-turbo": [
        {
          "name": "Stable Diffusion 3.5 (large_turbo)",
          "huggingface_id": "stabilityai/stable-diffusion-3.5-large-turbo",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "large_turbo"
        }
      ],
      "stepfun-ai/Step-3.5-Flash": [
        {
          "name": "Step-3.5-Flash",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-FP8": [
        {
          "name": "Step-3.5-Flash (fp8)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-INT4": [
        {
          "name": "Step-3.5-Flash (int4)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-INT4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-INT8": [
        {
          "name": "Step-3.5-Flash (int8)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-INT8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int8"
        }
      ],
      "stepfun-ai/Step-3.7-Flash": [
        {
          "name": "Step-3.7-Flash",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stepfun-ai/Step-3.7-Flash-FP8": [
        {
          "name": "Step-3.7-Flash (fp8)",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "stepfun-ai/Step-3.7-Flash-NVFP4": [
        {
          "name": "Step-3.7-Flash (nvfp4)",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--quantization": "modelopt",
                "--kv-cache-dtype": "fp8",
                "--gpu-memory-utilization": "0.9",
                "--async-scheduling": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--quantization": "modelopt",
                "--kv-cache-dtype": "fp8",
                "--gpu-memory-utilization": "0.9",
                "--async-scheduling": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "tencent/Hunyuan-A13B-Instruct": [
        {
          "name": "Hunyuan-A13B-Instruct",
          "huggingface_id": "tencent/Hunyuan-A13B-Instruct",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hunyuan-A13B-Instruct-FP8": [
        {
          "name": "Hunyuan-A13B-Instruct (fp8)",
          "huggingface_id": "tencent/Hunyuan-A13B-Instruct-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "tencent/HunyuanOCR": [
        {
          "name": "HunyuanOCR",
          "huggingface_id": "tencent/HunyuanOCR",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3-preview": [
        {
          "name": "Hy3-preview",
          "huggingface_id": "tencent/Hy3-preview",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "1",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3": [
        {
          "name": "Hy3",
          "huggingface_id": "tencent/Hy3",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3-FP8": [
        {
          "name": "Hy3 (fp8)",
          "huggingface_id": "tencent/Hy3-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_MOE": "0",
                "VLLM_ROCM_USE_AITER_MHA": "1",
                "VLLM_ROCM_USE_AITER_RMSNORM": "1",
                "VLLM_ROCM_USE_AITER_LINEAR": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "thinkingmachines/Inkling-NVFP4": [
        {
          "name": "TML Inkling",
          "huggingface_id": "thinkingmachines/Inkling-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "thinkingmachines/Inkling": [
        {
          "name": "TML Inkling (bf16)",
          "huggingface_id": "thinkingmachines/Inkling",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "zai-org/GLM-4.5": [
        {
          "name": "GLM-4.5",
          "huggingface_id": "zai-org/GLM-4.5",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.5-FP8": [
        {
          "name": "GLM-4.5 (fp8)",
          "huggingface_id": "zai-org/GLM-4.5-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.5V": [
        {
          "name": "GLM-4.5V",
          "huggingface_id": "zai-org/GLM-4.5V",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.5V-FP8": [
        {
          "name": "GLM-4.5V (fp8)",
          "huggingface_id": "zai-org/GLM-4.5V-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.6": [
        {
          "name": "GLM-4.6",
          "huggingface_id": "zai-org/GLM-4.6",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.6-FP8": [
        {
          "name": "GLM-4.6 (fp8)",
          "huggingface_id": "zai-org/GLM-4.6-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.6V": [
        {
          "name": "GLM-4.6V",
          "huggingface_id": "zai-org/GLM-4.6V",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.6V-FP8": [
        {
          "name": "GLM-4.6V (fp8)",
          "huggingface_id": "zai-org/GLM-4.6V-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "SAFETENSORS_FAST_GPU": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.7": [
        {
          "name": "GLM-4.7",
          "huggingface_id": "zai-org/GLM-4.7",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.7-FP8": [
        {
          "name": "GLM-4.7 (fp8)",
          "huggingface_id": "zai-org/GLM-4.7-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-4.7-NVFP4": [
        {
          "name": "GLM-4.7 (nvfp4)",
          "huggingface_id": "nvidia/GLM-4.7-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.1": [
        {
          "name": "GLM-5.1",
          "huggingface_id": "zai-org/GLM-5.1",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-5.1-FP8": [
        {
          "name": "GLM-5.1 (fp8)",
          "huggingface_id": "zai-org/GLM-5.1-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-5.1-NVFP4": [
        {
          "name": "GLM-5.1 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5.1-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.2-FP8": [
        {
          "name": "GLM-5.2",
          "huggingface_id": "zai-org/GLM-5.2-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/GLM-5.2-NVFP4": [
        {
          "name": "GLM-5.2 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5.2-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.2": [
        {
          "name": "GLM-5.2 (bf16)",
          "huggingface_id": "zai-org/GLM-5.2",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--linear-backend": "aiter",
                "--moe-backend": "aiter",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_ROCM_USE_AITER": "1",
                "VLLM_ROCM_USE_AITER_FUSION_SHARED_EXPERTS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "zai-org/GLM-5": [
        {
          "name": "GLM-5",
          "huggingface_id": "zai-org/GLM-5",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-5-FP8": [
        {
          "name": "GLM-5 (fp8)",
          "huggingface_id": "zai-org/GLM-5-FP8",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-5-NVFP4": [
        {
          "name": "GLM-5 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-ASR-Nano-2512": [
        {
          "name": "GLM-ASR-Nano-2512",
          "huggingface_id": "zai-org/GLM-ASR-Nano-2512",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-GA": [
        {
          "name": "GLM-GA",
          "huggingface_id": "zai-org/GLM-GA",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--allowed-local-media-path": "/",
                "--mm-processor-cache-type": "shm",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "glm4_6v"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-Image": [
        {
          "name": "GLM-Image",
          "huggingface_id": "zai-org/GLM-Image",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-OCR": [
        {
          "name": "GLM-OCR",
          "huggingface_id": "zai-org/GLM-OCR",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-TTS": [
        {
          "name": "GLM-TTS",
          "huggingface_id": "zai-org/GLM-TTS",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/Glyph": [
        {
          "name": "Glyph",
          "huggingface_id": "zai-org/Glyph",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--limit-mm-per-prompt.video": "0",
                "--reasoning-parser": "glm45",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-26B-A4B-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-26B-A4B-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-31B-IT-NVFP4": [
        {
          "name": "Gemma 4 31B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-31B-IT-NVFP4",
          "accelerators": "amd",
          "provider": "vllm_amd",
          "custom_image": "vllm/vllm-openai-rocm:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ]
    },
    "google": {
      "ByteDance-Seed/Seed-OSS-36B-Instruct": [
        {
          "name": "Seed-OSS-36B",
          "huggingface_id": "ByteDance-Seed/Seed-OSS-36B-Instruct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "seed_oss"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "google/diffusiongemma-26B-A4B-it": [
        {
          "name": "DiffusionGemma 26B-A4B IT",
          "huggingface_id": "google/diffusiongemma-26B-A4B-it",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/diffusiongemma-26B-A4B-it-FP8-dynamic": [
        {
          "name": "DiffusionGemma 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/diffusiongemma-26B-A4B-it-FP8-dynamic",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/diffusiongemma-26B-A4B-it-NVFP4": [
        {
          "name": "DiffusionGemma 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/diffusiongemma-26B-A4B-it-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-12B-it": [
        {
          "name": "Gemma 4 12B IT",
          "huggingface_id": "google/gemma-4-12B-it",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-12B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "google/gemma-4-12B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 12B IT (w4a16)",
          "huggingface_id": "google/gemma-4-12B-it-qat-w4a16-ct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-12B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-26B-A4B-it": [
        {
          "name": "Gemma 4 26B-A4B IT",
          "huggingface_id": "google/gemma-4-26B-A4B-it",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it": [
        {
          "name": "Gemma 4 31B IT",
          "huggingface_id": "google/gemma-4-31B-it",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-31B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 31B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-31B-it-FP8-dynamic",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/gemma-4-31B-it-NVFP4": [
        {
          "name": "Gemma 4 31B IT (nvfp4)",
          "huggingface_id": "nvidia/gemma-4-31B-it-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 31B IT (w4a16)",
          "huggingface_id": "google/gemma-4-31B-it-qat-w4a16-ct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it": [
        {
          "name": "Gemma 4 E2B IT",
          "huggingface_id": "google/gemma-4-E2B-it",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E2B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E2B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E2B-it-FP8-dynamic",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E2B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E2B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-w4a16-ct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E2B IT (mobile)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-mobile-ct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "google/gemma-4-E4B-it": [
        {
          "name": "Gemma 4 E4B IT",
          "huggingface_id": "google/gemma-4-E4B-it",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E4B-it-FP8-dynamic",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E4B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E4B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-w4a16-ct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E4B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E4B IT (mobile)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-mobile-ct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "google/translategemma-27b-it": [
        {
          "name": "TranslateGemma 27B IT",
          "huggingface_id": "google/translategemma-27b-it",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Infomaniak-AI/vllm-translategemma-27b-it": [
        {
          "name": "TranslateGemma 27B IT (vllm_optimized)",
          "huggingface_id": "Infomaniak-AI/vllm-translategemma-27b-it",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "vllm_optimized"
        }
      ],
      "google/translategemma-4b-it": [
        {
          "name": "TranslateGemma 27B IT (small_4b)",
          "huggingface_id": "google/translategemma-4b-it",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "small_4b"
        }
      ],
      "JetBrains/Mellum2-12B-A2.5B-Instruct": [
        {
          "name": "Mellum2-12B-A2.5B-Instruct",
          "huggingface_id": "JetBrains/Mellum2-12B-A2.5B-Instruct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "JetBrains/Mellum2-12B-A2.5B-Thinking": [
        {
          "name": "Mellum2-12B-A2.5B-Thinking",
          "huggingface_id": "JetBrains/Mellum2-12B-A2.5B-Thinking",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Base": [
        {
          "name": "LFM2.5 1.2B Base",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Base",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Instruct": [
        {
          "name": "LFM2.5 1.2B Instruct",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Instruct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-JP-202606": [
        {
          "name": "LFM2.5 1.2B JP (202606)",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-JP-202606",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-JP": [
        {
          "name": "LFM2.5 1.2B JP",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-JP",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Thinking": [
        {
          "name": "LFM2.5 1.2B Thinking",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Thinking",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-230M": [
        {
          "name": "LFM2.5 230M",
          "huggingface_id": "LiquidAI/LFM2.5-230M",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-350M": [
        {
          "name": "LFM2.5 350M",
          "huggingface_id": "LiquidAI/LFM2.5-350M",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-8B-A1B": [
        {
          "name": "LFM2.5 8B A1B",
          "huggingface_id": "LiquidAI/LFM2.5-8B-A1B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-VL-1.6B": [
        {
          "name": "LFM2.5 VL 1.6B",
          "huggingface_id": "LiquidAI/LFM2.5-VL-1.6B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-VL-450M": [
        {
          "name": "LFM2.5 VL 450M",
          "huggingface_id": "LiquidAI/LFM2.5-VL-450M",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M2.1": [
        {
          "name": "MiniMax-M2.1",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.1",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M2.5": [
        {
          "name": "MiniMax-M2.5",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.5",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.5-NVFP4": [
        {
          "name": "MiniMax-M2.5 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.5-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2.7": [
        {
          "name": "MiniMax-M2.7",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.7",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.7-NVFP4": [
        {
          "name": "MiniMax-M2.7 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.7-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2": [
        {
          "name": "MiniMax-M2",
          "huggingface_id": "MiniMaxAI/MiniMax-M2",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/MiniMax-M2-NVFP4": [
        {
          "name": "MiniMax-M2 (nvfp4)",
          "huggingface_id": "RedHatAI/MiniMax-M2-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M3": [
        {
          "name": "MiniMax-M3",
          "huggingface_id": "MiniMaxAI/MiniMax-M3",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M3-MXFP8": [
        {
          "name": "MiniMax-M3 (mxfp8)",
          "huggingface_id": "MiniMaxAI/MiniMax-M3-MXFP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp8"
        }
      ],
      "nvidia/MiniMax-M3-NVFP4": [
        {
          "name": "MiniMax-M3 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M3-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "amd/MiniMax-M3-MXFP4": [
        {
          "name": "MiniMax-M3 (mxfp4)",
          "huggingface_id": "amd/MiniMax-M3-MXFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp4"
        }
      ],
      "OpenGVLab/InternVL3_5-8B": [
        {
          "name": "InternVL3.5",
          "huggingface_id": "OpenGVLab/InternVL3_5-8B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-SoundEffect": [
        {
          "name": "MOSS-SoundEffect",
          "huggingface_id": "OpenMOSS-Team/MOSS-SoundEffect",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_sound_effect.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-TTS-Realtime": [
        {
          "name": "MOSS-TTS-Realtime",
          "huggingface_id": "OpenMOSS-Team/MOSS-TTS-Realtime",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-TTS": [
        {
          "name": "MOSS-TTS",
          "huggingface_id": "OpenMOSS-Team/MOSS-TTS",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_tts.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-TTSD-v1.0": [
        {
          "name": "MOSS-TTSD-v1.0",
          "huggingface_id": "OpenMOSS-Team/MOSS-TTSD-v1.0",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_ttsd.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-Transcribe-Diarize": [
        {
          "name": "MOSS-Transcribe-Diarize",
          "huggingface_id": "OpenMOSS-Team/MOSS-Transcribe-Diarize",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-VoiceGenerator": [
        {
          "name": "MOSS-VoiceGenerator",
          "huggingface_id": "OpenMOSS-Team/MOSS-VoiceGenerator",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_voice_generator.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "PaddlePaddle/PaddleOCR-VL-1.5": [
        {
          "name": "PaddleOCR-VL-1.5",
          "huggingface_id": "PaddlePaddle/PaddleOCR-VL-1.5",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-batched-tokens": "16384",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "PaddlePaddle/PaddleOCR-VL-1.6": [
        {
          "name": "PaddleOCR-VL-1.6",
          "huggingface_id": "PaddlePaddle/PaddleOCR-VL-1.6",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-batched-tokens": "16384",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "PaddlePaddle/PaddleOCR-VL": [
        {
          "name": "PaddleOCR-VL",
          "huggingface_id": "PaddlePaddle/PaddleOCR-VL",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-batched-tokens": "16384",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen-Image": [
        {
          "name": "Qwen-Image",
          "huggingface_id": "Qwen/Qwen-Image",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        },
        {
          "name": "Qwen-Image (fp8)",
          "huggingface_id": "Qwen/Qwen-Image",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "fp8",
                "--ignored-layers": "img_mlp"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        },
        {
          "name": "Qwen-Image (int8)",
          "huggingface_id": "Qwen/Qwen-Image",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "int8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int8"
        }
      ],
      "Qwen/Qwen2.5-VL-72B-Instruct": [
        {
          "name": "Qwen2.5-VL-72B-Instruct",
          "huggingface_id": "Qwen/Qwen2.5-VL-72B-Instruct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen2.5-VL-72B-Instruct-AWQ": [
        {
          "name": "Qwen2.5-VL-72B-Instruct (awq)",
          "huggingface_id": "Qwen/Qwen2.5-VL-72B-Instruct-AWQ",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "awq",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen2.5-VL-7B-Instruct": [
        {
          "name": "Qwen2.5-VL-7B-Instruct",
          "huggingface_id": "Qwen/Qwen2.5-VL-7B-Instruct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen2.5-VL-7B-Instruct-AWQ": [
        {
          "name": "Qwen2.5-VL-7B-Instruct (awq)",
          "huggingface_id": "Qwen/Qwen2.5-VL-7B-Instruct-AWQ",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "awq",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen3-235B-A22B-Instruct-2507": [
        {
          "name": "Qwen3-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-Instruct-2507",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-235B-A22B-FP8": [
        {
          "name": "Qwen3-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4": [
        {
          "name": "Qwen3-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-32B": [
        {
          "name": "Qwen3-32B",
          "huggingface_id": "Qwen/Qwen3-32B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-32B-FP8": [
        {
          "name": "Qwen3-32B (fp8)",
          "huggingface_id": "Qwen/Qwen3-32B-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3-32B-AWQ": [
        {
          "name": "Qwen3-32B (awq)",
          "huggingface_id": "Qwen/Qwen3-32B-AWQ",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "awq",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen3-4B": [
        {
          "name": "Qwen3-4B",
          "huggingface_id": "Qwen/Qwen3-4B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-4B-FP8": [
        {
          "name": "Qwen3-4B (fp8)",
          "huggingface_id": "Qwen/Qwen3-4B-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3-ASR-1.7B": [
        {
          "name": "Qwen3-ASR-1.7B",
          "huggingface_id": "Qwen/Qwen3-ASR-1.7B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (voicedesign-1.7b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "voicedesign-1.7b"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-1.7B-Base": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (base-1.7b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-1.7B-Base",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base-1.7b"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (customvoice-0.6b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "customvoice-0.6b"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-0.6B-Base": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (base-0.6b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-0.6B-Base",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base-0.6b"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-0.8B": [
        {
          "name": "Qwen3.5-0.8B",
          "huggingface_id": "Qwen/Qwen3.5-0.8B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-122B-A10B": [
        {
          "name": "Qwen3.5-122B-A10B",
          "huggingface_id": "Qwen/Qwen3.5-122B-A10B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-122B-A10B-FP8": [
        {
          "name": "Qwen3.5-122B-A10B (fp8)",
          "huggingface_id": "Qwen/Qwen3.5-122B-A10B-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3.5-122B-A10B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-122B-A10B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-122B-A10B-GPTQ-Int4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-27B": [
        {
          "name": "Qwen3.5-27B",
          "huggingface_id": "Qwen/Qwen3.5-27B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-27B-FP8": [
        {
          "name": "Qwen3.5-27B (fp8)",
          "huggingface_id": "Qwen/Qwen3.5-27B-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3.5-27B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-27B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-27B-GPTQ-Int4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-2B": [
        {
          "name": "Qwen3.5-2B",
          "huggingface_id": "Qwen/Qwen3.5-2B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-35B-A3B": [
        {
          "name": "Qwen3.5-35B-A3B",
          "huggingface_id": "Qwen/Qwen3.5-35B-A3B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-35B-A3B-FP8": [
        {
          "name": "Qwen3.5-35B-A3B (fp8)",
          "huggingface_id": "Qwen/Qwen3.5-35B-A3B-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3.5-35B-A3B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-35B-A3B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-35B-A3B-GPTQ-Int4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B": [
        {
          "name": "Qwen3.5-397B",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Qwen3.5-397B-A17B-NVFP4": [
        {
          "name": "Qwen3.5-397B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.5-397B-A17B-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-397B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-4B": [
        {
          "name": "Qwen3.5-4B",
          "huggingface_id": "Qwen/Qwen3.5-4B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-9B": [
        {
          "name": "Qwen3.5-9B",
          "huggingface_id": "Qwen/Qwen3.5-9B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-27B": [
        {
          "name": "Qwen3.6-27B",
          "huggingface_id": "Qwen/Qwen3.6-27B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-27B-FP8": [
        {
          "name": "Qwen3.6-27B (fp8)",
          "huggingface_id": "Qwen/Qwen3.6-27B-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "512",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3.6-27B-NVFP4": [
        {
          "name": "Qwen3.6-27B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.6-27B-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B": [
        {
          "name": "Qwen3.6-35B-A3B",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B-FP8": [
        {
          "name": "Qwen3.6-35B-A3B (fp8)",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3.6-35B-A3B-NVFP4": [
        {
          "name": "Qwen3.6-35B-A3B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.6-35B-A3B-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3Guard-Gen-8B": [
        {
          "name": "Qwen3Guard-Gen-8B",
          "huggingface_id": "Qwen/Qwen3Guard-Gen-8B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3Guard-Gen-4B": [
        {
          "name": "Qwen3Guard-Gen-8B (small_4b)",
          "huggingface_id": "Qwen/Qwen3Guard-Gen-4B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "small_4b"
        }
      ],
      "Qwen/Qwen3Guard-Gen-0.6B": [
        {
          "name": "Qwen3Guard-Gen-8B (tiny_0_6b)",
          "huggingface_id": "Qwen/Qwen3Guard-Gen-0.6B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "tiny_0_6b"
        }
      ],
      "Wan-AI/Wan2.2-T2V-A14B-Diffusers": [
        {
          "name": "Wan2.2",
          "huggingface_id": "Wan-AI/Wan2.2-T2V-A14B-Diffusers",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "XiaomiMiMo/MiMo-V2-Flash": [
        {
          "name": "MiMo-V2-Flash",
          "huggingface_id": "XiaomiMiMo/MiMo-V2-Flash",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "XiaomiMiMo/MiMo-V2.5-Pro": [
        {
          "name": "MiMo-V2.5-Pro",
          "huggingface_id": "XiaomiMiMo/MiMo-V2.5-Pro",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "XiaomiMiMo/MiMo-V2.5": [
        {
          "name": "MiMo-V2.5",
          "huggingface_id": "XiaomiMiMo/MiMo-V2.5",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "arcee-ai/Trinity-Large-Thinking": [
        {
          "name": "Trinity-Large-Thinking",
          "huggingface_id": "arcee-ai/Trinity-Large-Thinking",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "arcee-ai/Trinity-Large-Thinking-NVFP4": [
        {
          "name": "Trinity-Large-Thinking (nvfp4)",
          "huggingface_id": "arcee-ai/Trinity-Large-Thinking-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "baidu/ERNIE-4.5-21B-A3B-PT": [
        {
          "name": "ERNIE-4.5",
          "huggingface_id": "baidu/ERNIE-4.5-21B-A3B-PT",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "baidu/ERNIE-4.5-300B-A47B-PT": [
        {
          "name": "ERNIE-4.5 (300b)",
          "huggingface_id": "baidu/ERNIE-4.5-300B-A47B-PT",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--gpu-memory-utilization": "0.95",
                "--quantization": "fp8",
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--gpu-memory-utilization": "0.95",
                "--quantization": "fp8",
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "300b"
        }
      ],
      "baidu/ERNIE-4.5-VL-28B-A3B-PT": [
        {
          "name": "ERNIE-4.5-VL",
          "huggingface_id": "baidu/ERNIE-4.5-VL-28B-A3B-PT",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "baidu/ERNIE-4.5-VL-424B-A47B-PT": [
        {
          "name": "ERNIE-4.5-VL (424b)",
          "huggingface_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "424b"
        },
        {
          "name": "ERNIE-4.5-VL (424b_fp8)",
          "huggingface_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--quantization": "fp8",
                "--cpu-offload-gb": "50",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--quantization": "fp8",
                "--cpu-offload-gb": "50",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "424b_fp8"
        }
      ],
      "baidu/Unlimited-OCR": [
        {
          "name": "Unlimited-OCR",
          "huggingface_id": "baidu/Unlimited-OCR",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--logits_processors": "vllm.model_executor.models.unlimited_ocr:NGramPerReqLogitsProcessor",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "bosonai/higgs-audio-v3-tts-4b": [
        {
          "name": "Higgs Audio V3 TTS",
          "huggingface_id": "bosonai/higgs-audio-v3-tts-4b",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-OCR-2": [
        {
          "name": "DeepSeek-OCR-2",
          "huggingface_id": "deepseek-ai/DeepSeek-OCR-2",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--logits_processors": "vllm.model_executor.models.deepseek_ocr:NGramPerReqLogitsProcessor",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-OCR": [
        {
          "name": "DeepSeek-OCR",
          "huggingface_id": "deepseek-ai/DeepSeek-OCR",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--logits_processors": "vllm.model_executor.models.deepseek_ocr:NGramPerReqLogitsProcessor",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-R1": [
        {
          "name": "DeepSeek-R1",
          "huggingface_id": "deepseek-ai/DeepSeek-R1",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-R1-0528": [
        {
          "name": "DeepSeek-R1 (r1_0528)",
          "huggingface_id": "deepseek-ai/DeepSeek-R1-0528",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "r1_0528"
        }
      ],
      "nvidia/DeepSeek-R1-0528-NVFP4-v2": [
        {
          "name": "DeepSeek-R1 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-R1-0528-NVFP4-v2",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3.1": [
        {
          "name": "DeepSeek-V3.1",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.1",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3.1-NVFP4": [
        {
          "name": "DeepSeek-V3.1 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V3.1-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3.2-Exp": [
        {
          "name": "DeepSeek-V3.2-Exp",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.2-Exp",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-V3.2": [
        {
          "name": "DeepSeek-V3.2",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.2",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3.2-NVFP4": [
        {
          "name": "DeepSeek-V3.2 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V3.2-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--kv-cache-dtype": "fp8",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--kv-cache-dtype": "fp8",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3": [
        {
          "name": "DeepSeek-V3",
          "huggingface_id": "deepseek-ai/DeepSeek-V3",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3-FP4": [
        {
          "name": "DeepSeek-V3 (fp4)",
          "huggingface_id": "nvidia/DeepSeek-V3-FP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Flash": [
        {
          "name": "DeepSeek-V4-Flash",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Flash",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Flash-NVFP4": [
        {
          "name": "DeepSeek-V4-Flash (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Flash-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro": [
        {
          "name": "DeepSeek-V4-Pro",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Pro-NVFP4": [
        {
          "name": "DeepSeek-V4-Pro (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Pro-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro-DSpark": [
        {
          "name": "DeepSeek-V4-Pro (dspark)",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro-DSpark",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "dspark"
        }
      ],
      "fishaudio/s2-pro": [
        {
          "name": "Fish Speech S2 Pro",
          "huggingface_id": "fishaudio/s2-pro",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ling-2.6-1T": [
        {
          "name": "Ling-2.6-1T",
          "huggingface_id": "inclusionAI/Ling-2.6-1T",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ling-2.6-flash": [
        {
          "name": "Ling-2.6-flash",
          "huggingface_id": "inclusionAI/Ling-2.6-flash",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ming-omni-tts-0.5B": [
        {
          "name": "Ming-omni-tts-0.5B",
          "huggingface_id": "inclusionAI/Ming-omni-tts-0.5B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--enforce-eager": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ring-1T-FP8": [
        {
          "name": "Ring-1T-FP8",
          "huggingface_id": "inclusionAI/Ring-1T-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ring-2.6-1T": [
        {
          "name": "Ring-2.6-1T",
          "huggingface_id": "inclusionAI/Ring-2.6-1T",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S1": [
        {
          "name": "Intern-S1",
          "huggingface_id": "internlm/Intern-S1",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S1-FP8": [
        {
          "name": "Intern-S1 (fp8)",
          "huggingface_id": "internlm/Intern-S1-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "internlm/Intern-S2-Preview": [
        {
          "name": "Intern-S2-Preview",
          "huggingface_id": "internlm/Intern-S2-Preview",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S2-Preview-FP8": [
        {
          "name": "Intern-S2-Preview (fp8)",
          "huggingface_id": "internlm/Intern-S2-Preview-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-retrieval": [
        {
          "name": "Jina Embeddings v5 Text Small",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-retrieval",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-text-matching": [
        {
          "name": "Jina Embeddings v5 Text Small (text_matching)",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-text-matching",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "text_matching"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-classification": [
        {
          "name": "Jina Embeddings v5 Text Small (classification)",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-classification",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "classification"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-clustering": [
        {
          "name": "Jina Embeddings v5 Text Small (clustering)",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-clustering",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "clustering"
        }
      ],
      "jinaai/jina-reranker-m0": [
        {
          "name": "Jina Reranker m0",
          "huggingface_id": "jinaai/jina-reranker-m0",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--gpu-memory-utilization": "0.75",
                "--max-num-seqs": "32"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "meituan-longcat/LongCat-Image-Edit": [
        {
          "name": "LongCat-Image-Edit",
          "huggingface_id": "meituan-longcat/LongCat-Image-Edit",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "meta-llama/Llama-3.3-70B-Instruct": [
        {
          "name": "Llama-3.3-70B",
          "huggingface_id": "meta-llama/Llama-3.3-70B-Instruct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-FP8": [
        {
          "name": "Llama-3.3-70B (fp8)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-NVFP4": [
        {
          "name": "Llama-3.3-70B (nvfp4)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "meta-llama/Llama-4-Scout-17B-16E-Instruct": [
        {
          "name": "Llama-4-Scout",
          "huggingface_id": "meta-llama/Llama-4-Scout-17B-16E-Instruct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8": [
        {
          "name": "Llama-4-Scout (fp8)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4": [
        {
          "name": "Llama-4-Scout (nvfp4)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "microsoft/Phi-4-mini-instruct": [
        {
          "name": "Phi-4",
          "huggingface_id": "microsoft/Phi-4-mini-instruct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "microsoft/Phi-4-mini-reasoning": [
        {
          "name": "Phi-4 (mini_reasoning)",
          "huggingface_id": "microsoft/Phi-4-mini-reasoning",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mini_reasoning"
        }
      ],
      "microsoft/Phi-4-reasoning": [
        {
          "name": "Phi-4 (reasoning)",
          "huggingface_id": "microsoft/Phi-4-reasoning",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "reasoning"
        }
      ],
      "microsoft/Phi-4-multimodal-instruct": [
        {
          "name": "Phi-4 (multimodal)",
          "huggingface_id": "microsoft/Phi-4-multimodal-instruct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "multimodal"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Ministral-3-8B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (8b)",
          "huggingface_id": "mistralai/Ministral-3-8B-Instruct-2512",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "8b"
        }
      ],
      "mistralai/Ministral-3-3B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (3b)",
          "huggingface_id": "mistralai/Ministral-3-3B-Instruct-2512",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "3b"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512-FP8": [
        {
          "name": "Ministral-3-Instruct (fp8)",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "mistralai/Ministral-3-8B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning",
          "huggingface_id": "mistralai/Ministral-3-8B-Reasoning-2512",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Ministral-3-3B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning (3b)",
          "huggingface_id": "mistralai/Ministral-3-3B-Reasoning-2512",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "3b"
        }
      ],
      "mistralai/Ministral-3-14B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning (14b)",
          "huggingface_id": "mistralai/Ministral-3-14B-Reasoning-2512",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "14b"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512": [
        {
          "name": "Mistral-Large-3-675B-Instruct",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4": [
        {
          "name": "Mistral-Large-3-675B-Instruct (nvfp4)",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512-FP8": [
        {
          "name": "Mistral-Large-3-675B-Instruct (fp8)",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "mistralai/Mistral-Medium-3.5-128B": [
        {
          "name": "Mistral-Medium-3.5",
          "huggingface_id": "mistralai/Mistral-Medium-3.5-128B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Medium-3.5-128B-EAGLE\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Mistral-Small-4-119B-2603": [
        {
          "name": "Mistral-Small-4-119B",
          "huggingface_id": "mistralai/Mistral-Small-4-119B-2603",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "FLASH_ATTN_MLA",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "FLASH_ATTN_MLA",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Mistral-Small-4-119B-2603-NVFP4": [
        {
          "name": "Mistral-Small-4-119B (nvfp4)",
          "huggingface_id": "mistralai/Mistral-Small-4-119B-2603-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "TRITON_MLA",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "TRITON_MLA",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "mistralai/Voxtral-4B-TTS-2603": [
        {
          "name": "Voxtral-4B-TTS-2603",
          "huggingface_id": "mistralai/Voxtral-4B-TTS-2603",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Voxtral-Mini-4B-Realtime-2602": [
        {
          "name": "Voxtral-Mini-4B-Realtime-2602",
          "huggingface_id": "mistralai/Voxtral-Mini-4B-Realtime-2602",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer-mode": "mistral",
                "--compilation_config": "{\"cudagraph_mode\": \"PIECEWISE\"}"
              },
              "extra_envs": {
                "VLLM_DISABLE_COMPILE_CACHE": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-K2-Instruct": [
        {
          "name": "Kimi-K2-Instruct",
          "huggingface_id": "moonshotai/Kimi-K2-Instruct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-K2-Thinking": [
        {
          "name": "Kimi-K2-Thinking",
          "huggingface_id": "moonshotai/Kimi-K2-Thinking",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2-Thinking-NVFP4": [
        {
          "name": "Kimi-K2-Thinking (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2-Thinking-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.5": [
        {
          "name": "Kimi-K2.5",
          "huggingface_id": "moonshotai/Kimi-K2.5",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.5-NVFP4": [
        {
          "name": "Kimi-K2.5 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.5-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.6": [
        {
          "name": "Kimi-K2.6",
          "huggingface_id": "moonshotai/Kimi-K2.6",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.6-NVFP4": [
        {
          "name": "Kimi-K2.6 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.6-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.7-Code": [
        {
          "name": "Kimi-K2.7-Code",
          "huggingface_id": "moonshotai/Kimi-K2.7-Code",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-Linear-48B-A3B-Instruct": [
        {
          "name": "Kimi-Linear-48B-A3B-Instruct",
          "huggingface_id": "moonshotai/Kimi-Linear-48B-A3B-Instruct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Nano": [
        {
          "name": "Cosmos3-Nano",
          "huggingface_id": "nvidia/Cosmos3-Nano",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super-Image2Video": [
        {
          "name": "Cosmos3-Super-Image2Video",
          "huggingface_id": "nvidia/Cosmos3-Super-Image2Video",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super-Text2Image": [
        {
          "name": "Cosmos3-Super-Text2Image",
          "huggingface_id": "nvidia/Cosmos3-Super-Text2Image",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super": [
        {
          "name": "Cosmos3-Super",
          "huggingface_id": "nvidia/Cosmos3-Super",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Nano-30B-A3B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Nano-30B-A3B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_FLASHINFER_MOE_BACKEND": "throughput"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_FLASHINFER_MOE_BACKEND": "throughput"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Nano-4B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Nano-4B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (base_bf16)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base_bf16"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Ultra-550B-A55B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}"
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4": [
        {
          "name": "NVIDIA Nemotron-3-Ultra-550B-A55B (fp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}"
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-FP8": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-NVFP4-QAD": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-NVFP4-QAD",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-FP8": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-NVFP4": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Base": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (base)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Base",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Japanese": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (japanese)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Japanese",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "japanese"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning (fp8)",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning (nvfp4)",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "openai/gpt-oss-120b": [
        {
          "name": "GPT-OSS",
          "huggingface_id": "openai/gpt-oss-120b",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "amd/gpt-oss-120b-w-mxfp4-a-fp8": [
        {
          "name": "GPT-OSS (amd_fp8)",
          "huggingface_id": "amd/gpt-oss-120b-w-mxfp4-a-fp8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "amd_fp8"
        }
      ],
      "openai/gpt-oss-20b": [
        {
          "name": "GPT-OSS 20B",
          "huggingface_id": "openai/gpt-oss-20b",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "openbmb/MiniCPM-V-4.6": [
        {
          "name": "MiniCPM-V 4.6",
          "huggingface_id": "openbmb/MiniCPM-V-4.6",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "openbmb/MiniCPM5-1B": [
        {
          "name": "MiniCPM5-1B",
          "huggingface_id": "openbmb/MiniCPM5-1B",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "openbmb/VoxCPM2": [
        {
          "name": "VoxCPM2",
          "huggingface_id": "openbmb/VoxCPM2",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "pfnet/plamo-2-translate": [
        {
          "name": "PLaMo 2 Translate",
          "huggingface_id": "pfnet/plamo-2-translate",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "pfnet/plamo-3-nict-31b-base": [
        {
          "name": "PLaMo 3 NICT 31B Base",
          "huggingface_id": "pfnet/plamo-3-nict-31b-base",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-M.1": [
        {
          "name": "Laguna M.1",
          "huggingface_id": "poolside/Laguna-M.1",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-M.1-FP8": [
        {
          "name": "Laguna M.1 (fp8)",
          "huggingface_id": "poolside/Laguna-M.1-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "poolside/Laguna-M.1-NVFP4": [
        {
          "name": "Laguna M.1 (nvfp4)",
          "huggingface_id": "poolside/Laguna-M.1-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "poolside/Laguna-XS-2.1": [
        {
          "name": "Laguna XS-2.1",
          "huggingface_id": "poolside/Laguna-XS-2.1",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-XS-2.1-FP8": [
        {
          "name": "Laguna XS-2.1 (fp8)",
          "huggingface_id": "poolside/Laguna-XS-2.1-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "poolside/Laguna-XS-2.1-NVFP4": [
        {
          "name": "Laguna XS-2.1 (nvfp4)",
          "huggingface_id": "poolside/Laguna-XS-2.1-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "poolside/Laguna-XS-2.1-INT4": [
        {
          "name": "Laguna XS-2.1 (int4)",
          "huggingface_id": "poolside/Laguna-XS-2.1-INT4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "poolside/Laguna-XS.2": [
        {
          "name": "Laguna XS.2",
          "huggingface_id": "poolside/Laguna-XS.2",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-XS.2-FP8": [
        {
          "name": "Laguna XS.2 (fp8)",
          "huggingface_id": "poolside/Laguna-XS.2-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "poolside/Laguna-XS.2-NVFP4": [
        {
          "name": "Laguna XS.2 (nvfp4)",
          "huggingface_id": "poolside/Laguna-XS.2-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "poolside/Laguna-XS.2-INT4": [
        {
          "name": "Laguna XS.2 (int4)",
          "huggingface_id": "poolside/Laguna-XS.2-INT4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "stabilityai/stable-audio-open-1.0": [
        {
          "name": "Stable Audio Open",
          "huggingface_id": "stabilityai/stable-audio-open-1.0",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enforce-eager": "",
                "--gpu-memory-utilization": "0.9"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stabilityai/stable-diffusion-3.5-medium": [
        {
          "name": "Stable Diffusion 3.5",
          "huggingface_id": "stabilityai/stable-diffusion-3.5-medium",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stabilityai/stable-diffusion-3.5-large": [
        {
          "name": "Stable Diffusion 3.5 (large)",
          "huggingface_id": "stabilityai/stable-diffusion-3.5-large",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "large"
        }
      ],
      "stabilityai/stable-diffusion-3.5-large-turbo": [
        {
          "name": "Stable Diffusion 3.5 (large_turbo)",
          "huggingface_id": "stabilityai/stable-diffusion-3.5-large-turbo",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "large_turbo"
        }
      ],
      "stepfun-ai/Step-3.5-Flash": [
        {
          "name": "Step-3.5-Flash",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-FP8": [
        {
          "name": "Step-3.5-Flash (fp8)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-INT4": [
        {
          "name": "Step-3.5-Flash (int4)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-INT4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-INT8": [
        {
          "name": "Step-3.5-Flash (int8)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-INT8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int8"
        }
      ],
      "stepfun-ai/Step-3.7-Flash": [
        {
          "name": "Step-3.7-Flash",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stepfun-ai/Step-3.7-Flash-FP8": [
        {
          "name": "Step-3.7-Flash (fp8)",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "stepfun-ai/Step-3.7-Flash-NVFP4": [
        {
          "name": "Step-3.7-Flash (nvfp4)",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--quantization": "modelopt",
                "--kv-cache-dtype": "fp8",
                "--gpu-memory-utilization": "0.9",
                "--async-scheduling": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--quantization": "modelopt",
                "--kv-cache-dtype": "fp8",
                "--gpu-memory-utilization": "0.9",
                "--async-scheduling": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "tencent/Hunyuan-A13B-Instruct": [
        {
          "name": "Hunyuan-A13B-Instruct",
          "huggingface_id": "tencent/Hunyuan-A13B-Instruct",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hunyuan-A13B-Instruct-FP8": [
        {
          "name": "Hunyuan-A13B-Instruct (fp8)",
          "huggingface_id": "tencent/Hunyuan-A13B-Instruct-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "tencent/HunyuanOCR": [
        {
          "name": "HunyuanOCR",
          "huggingface_id": "tencent/HunyuanOCR",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3-preview": [
        {
          "name": "Hy3-preview",
          "huggingface_id": "tencent/Hy3-preview",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3": [
        {
          "name": "Hy3",
          "huggingface_id": "tencent/Hy3",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3-FP8": [
        {
          "name": "Hy3 (fp8)",
          "huggingface_id": "tencent/Hy3-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "thinkingmachines/Inkling-NVFP4": [
        {
          "name": "TML Inkling",
          "huggingface_id": "thinkingmachines/Inkling-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "thinkingmachines/Inkling": [
        {
          "name": "TML Inkling (bf16)",
          "huggingface_id": "thinkingmachines/Inkling",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "zai-org/GLM-4.5": [
        {
          "name": "GLM-4.5",
          "huggingface_id": "zai-org/GLM-4.5",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.5-FP8": [
        {
          "name": "GLM-4.5 (fp8)",
          "huggingface_id": "zai-org/GLM-4.5-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.5V": [
        {
          "name": "GLM-4.5V",
          "huggingface_id": "zai-org/GLM-4.5V",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.5V-FP8": [
        {
          "name": "GLM-4.5V (fp8)",
          "huggingface_id": "zai-org/GLM-4.5V-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.6": [
        {
          "name": "GLM-4.6",
          "huggingface_id": "zai-org/GLM-4.6",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.6-FP8": [
        {
          "name": "GLM-4.6 (fp8)",
          "huggingface_id": "zai-org/GLM-4.6-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.6V": [
        {
          "name": "GLM-4.6V",
          "huggingface_id": "zai-org/GLM-4.6V",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.6V-FP8": [
        {
          "name": "GLM-4.6V (fp8)",
          "huggingface_id": "zai-org/GLM-4.6V-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.7": [
        {
          "name": "GLM-4.7",
          "huggingface_id": "zai-org/GLM-4.7",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.7-FP8": [
        {
          "name": "GLM-4.7 (fp8)",
          "huggingface_id": "zai-org/GLM-4.7-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-4.7-NVFP4": [
        {
          "name": "GLM-4.7 (nvfp4)",
          "huggingface_id": "nvidia/GLM-4.7-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.1": [
        {
          "name": "GLM-5.1",
          "huggingface_id": "zai-org/GLM-5.1",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-5.1-FP8": [
        {
          "name": "GLM-5.1 (fp8)",
          "huggingface_id": "zai-org/GLM-5.1-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-5.1-NVFP4": [
        {
          "name": "GLM-5.1 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5.1-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.2-FP8": [
        {
          "name": "GLM-5.2",
          "huggingface_id": "zai-org/GLM-5.2-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/GLM-5.2-NVFP4": [
        {
          "name": "GLM-5.2 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5.2-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.2": [
        {
          "name": "GLM-5.2 (bf16)",
          "huggingface_id": "zai-org/GLM-5.2",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "zai-org/GLM-5": [
        {
          "name": "GLM-5",
          "huggingface_id": "zai-org/GLM-5",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-5-FP8": [
        {
          "name": "GLM-5 (fp8)",
          "huggingface_id": "zai-org/GLM-5-FP8",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-5-NVFP4": [
        {
          "name": "GLM-5 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-ASR-Nano-2512": [
        {
          "name": "GLM-ASR-Nano-2512",
          "huggingface_id": "zai-org/GLM-ASR-Nano-2512",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-GA": [
        {
          "name": "GLM-GA",
          "huggingface_id": "zai-org/GLM-GA",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--allowed-local-media-path": "/",
                "--mm-processor-cache-type": "shm",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "glm4_6v"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-Image": [
        {
          "name": "GLM-Image",
          "huggingface_id": "zai-org/GLM-Image",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-OCR": [
        {
          "name": "GLM-OCR",
          "huggingface_id": "zai-org/GLM-OCR",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-TTS": [
        {
          "name": "GLM-TTS",
          "huggingface_id": "zai-org/GLM-TTS",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/Glyph": [
        {
          "name": "Glyph",
          "huggingface_id": "zai-org/Glyph",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--limit-mm-per-prompt.video": "0",
                "--reasoning-parser": "glm45",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-26B-A4B-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-26B-A4B-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-31B-IT-NVFP4": [
        {
          "name": "Gemma 4 31B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-31B-IT-NVFP4",
          "accelerators": "google",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ]
    },
    "intel": {
      "ByteDance-Seed/Seed-OSS-36B-Instruct": [
        {
          "name": "Seed-OSS-36B",
          "huggingface_id": "ByteDance-Seed/Seed-OSS-36B-Instruct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "seed_oss"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "google/diffusiongemma-26B-A4B-it": [
        {
          "name": "DiffusionGemma 26B-A4B IT",
          "huggingface_id": "google/diffusiongemma-26B-A4B-it",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/diffusiongemma-26B-A4B-it-FP8-dynamic": [
        {
          "name": "DiffusionGemma 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/diffusiongemma-26B-A4B-it-FP8-dynamic",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/diffusiongemma-26B-A4B-it-NVFP4": [
        {
          "name": "DiffusionGemma 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/diffusiongemma-26B-A4B-it-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-12B-it": [
        {
          "name": "Gemma 4 12B IT",
          "huggingface_id": "google/gemma-4-12B-it",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-12B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "google/gemma-4-12B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 12B IT (w4a16)",
          "huggingface_id": "google/gemma-4-12B-it-qat-w4a16-ct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-12B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-26B-A4B-it": [
        {
          "name": "Gemma 4 26B-A4B IT",
          "huggingface_id": "google/gemma-4-26B-A4B-it",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it": [
        {
          "name": "Gemma 4 31B IT",
          "huggingface_id": "google/gemma-4-31B-it",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-31B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 31B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-31B-it-FP8-dynamic",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/gemma-4-31B-it-NVFP4": [
        {
          "name": "Gemma 4 31B IT (nvfp4)",
          "huggingface_id": "nvidia/gemma-4-31B-it-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 31B IT (w4a16)",
          "huggingface_id": "google/gemma-4-31B-it-qat-w4a16-ct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it": [
        {
          "name": "Gemma 4 E2B IT",
          "huggingface_id": "google/gemma-4-E2B-it",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E2B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E2B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E2B-it-FP8-dynamic",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E2B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E2B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-w4a16-ct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E2B IT (mobile)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-mobile-ct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "google/gemma-4-E4B-it": [
        {
          "name": "Gemma 4 E4B IT",
          "huggingface_id": "google/gemma-4-E4B-it",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E4B-it-FP8-dynamic",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E4B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E4B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-w4a16-ct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E4B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E4B IT (mobile)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-mobile-ct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "google/translategemma-27b-it": [
        {
          "name": "TranslateGemma 27B IT",
          "huggingface_id": "google/translategemma-27b-it",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Infomaniak-AI/vllm-translategemma-27b-it": [
        {
          "name": "TranslateGemma 27B IT (vllm_optimized)",
          "huggingface_id": "Infomaniak-AI/vllm-translategemma-27b-it",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "vllm_optimized"
        }
      ],
      "google/translategemma-4b-it": [
        {
          "name": "TranslateGemma 27B IT (small_4b)",
          "huggingface_id": "google/translategemma-4b-it",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "small_4b"
        }
      ],
      "JetBrains/Mellum2-12B-A2.5B-Instruct": [
        {
          "name": "Mellum2-12B-A2.5B-Instruct",
          "huggingface_id": "JetBrains/Mellum2-12B-A2.5B-Instruct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "JetBrains/Mellum2-12B-A2.5B-Thinking": [
        {
          "name": "Mellum2-12B-A2.5B-Thinking",
          "huggingface_id": "JetBrains/Mellum2-12B-A2.5B-Thinking",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Base": [
        {
          "name": "LFM2.5 1.2B Base",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Base",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Instruct": [
        {
          "name": "LFM2.5 1.2B Instruct",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Instruct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-JP-202606": [
        {
          "name": "LFM2.5 1.2B JP (202606)",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-JP-202606",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-JP": [
        {
          "name": "LFM2.5 1.2B JP",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-JP",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Thinking": [
        {
          "name": "LFM2.5 1.2B Thinking",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Thinking",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-230M": [
        {
          "name": "LFM2.5 230M",
          "huggingface_id": "LiquidAI/LFM2.5-230M",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-350M": [
        {
          "name": "LFM2.5 350M",
          "huggingface_id": "LiquidAI/LFM2.5-350M",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-8B-A1B": [
        {
          "name": "LFM2.5 8B A1B",
          "huggingface_id": "LiquidAI/LFM2.5-8B-A1B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-VL-1.6B": [
        {
          "name": "LFM2.5 VL 1.6B",
          "huggingface_id": "LiquidAI/LFM2.5-VL-1.6B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-VL-450M": [
        {
          "name": "LFM2.5 VL 450M",
          "huggingface_id": "LiquidAI/LFM2.5-VL-450M",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M2.1": [
        {
          "name": "MiniMax-M2.1",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.1",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M2.5": [
        {
          "name": "MiniMax-M2.5",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.5",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.5-NVFP4": [
        {
          "name": "MiniMax-M2.5 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.5-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2.7": [
        {
          "name": "MiniMax-M2.7",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.7",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.7-NVFP4": [
        {
          "name": "MiniMax-M2.7 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.7-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2": [
        {
          "name": "MiniMax-M2",
          "huggingface_id": "MiniMaxAI/MiniMax-M2",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/MiniMax-M2-NVFP4": [
        {
          "name": "MiniMax-M2 (nvfp4)",
          "huggingface_id": "RedHatAI/MiniMax-M2-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M3": [
        {
          "name": "MiniMax-M3",
          "huggingface_id": "MiniMaxAI/MiniMax-M3",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M3-MXFP8": [
        {
          "name": "MiniMax-M3 (mxfp8)",
          "huggingface_id": "MiniMaxAI/MiniMax-M3-MXFP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp8"
        }
      ],
      "nvidia/MiniMax-M3-NVFP4": [
        {
          "name": "MiniMax-M3 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M3-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "amd/MiniMax-M3-MXFP4": [
        {
          "name": "MiniMax-M3 (mxfp4)",
          "huggingface_id": "amd/MiniMax-M3-MXFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp4"
        }
      ],
      "OpenGVLab/InternVL3_5-8B": [
        {
          "name": "InternVL3.5",
          "huggingface_id": "OpenGVLab/InternVL3_5-8B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-SoundEffect": [
        {
          "name": "MOSS-SoundEffect",
          "huggingface_id": "OpenMOSS-Team/MOSS-SoundEffect",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_sound_effect.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-TTS-Realtime": [
        {
          "name": "MOSS-TTS-Realtime",
          "huggingface_id": "OpenMOSS-Team/MOSS-TTS-Realtime",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-TTS": [
        {
          "name": "MOSS-TTS",
          "huggingface_id": "OpenMOSS-Team/MOSS-TTS",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_tts.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-TTSD-v1.0": [
        {
          "name": "MOSS-TTSD-v1.0",
          "huggingface_id": "OpenMOSS-Team/MOSS-TTSD-v1.0",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_ttsd.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-Transcribe-Diarize": [
        {
          "name": "MOSS-Transcribe-Diarize",
          "huggingface_id": "OpenMOSS-Team/MOSS-Transcribe-Diarize",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-VoiceGenerator": [
        {
          "name": "MOSS-VoiceGenerator",
          "huggingface_id": "OpenMOSS-Team/MOSS-VoiceGenerator",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_voice_generator.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "PaddlePaddle/PaddleOCR-VL-1.5": [
        {
          "name": "PaddleOCR-VL-1.5",
          "huggingface_id": "PaddlePaddle/PaddleOCR-VL-1.5",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-batched-tokens": "16384",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "PaddlePaddle/PaddleOCR-VL-1.6": [
        {
          "name": "PaddleOCR-VL-1.6",
          "huggingface_id": "PaddlePaddle/PaddleOCR-VL-1.6",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-batched-tokens": "16384",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "PaddlePaddle/PaddleOCR-VL": [
        {
          "name": "PaddleOCR-VL",
          "huggingface_id": "PaddlePaddle/PaddleOCR-VL",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-batched-tokens": "16384",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen-Image": [
        {
          "name": "Qwen-Image",
          "huggingface_id": "Qwen/Qwen-Image",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        },
        {
          "name": "Qwen-Image (fp8)",
          "huggingface_id": "Qwen/Qwen-Image",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "fp8",
                "--ignored-layers": "img_mlp"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        },
        {
          "name": "Qwen-Image (int8)",
          "huggingface_id": "Qwen/Qwen-Image",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "int8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int8"
        }
      ],
      "Qwen/Qwen2.5-VL-72B-Instruct": [
        {
          "name": "Qwen2.5-VL-72B-Instruct",
          "huggingface_id": "Qwen/Qwen2.5-VL-72B-Instruct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen2.5-VL-72B-Instruct-AWQ": [
        {
          "name": "Qwen2.5-VL-72B-Instruct (awq)",
          "huggingface_id": "Qwen/Qwen2.5-VL-72B-Instruct-AWQ",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "awq",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen2.5-VL-7B-Instruct": [
        {
          "name": "Qwen2.5-VL-7B-Instruct",
          "huggingface_id": "Qwen/Qwen2.5-VL-7B-Instruct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen2.5-VL-7B-Instruct-AWQ": [
        {
          "name": "Qwen2.5-VL-7B-Instruct (awq)",
          "huggingface_id": "Qwen/Qwen2.5-VL-7B-Instruct-AWQ",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "awq",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen3-235B-A22B-Instruct-2507": [
        {
          "name": "Qwen3-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-Instruct-2507",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-235B-A22B-FP8": [
        {
          "name": "Qwen3-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4": [
        {
          "name": "Qwen3-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-32B": [
        {
          "name": "Qwen3-32B",
          "huggingface_id": "Qwen/Qwen3-32B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-32B-FP8": [
        {
          "name": "Qwen3-32B (fp8)",
          "huggingface_id": "Qwen/Qwen3-32B-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3-32B-AWQ": [
        {
          "name": "Qwen3-32B (awq)",
          "huggingface_id": "Qwen/Qwen3-32B-AWQ",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "awq",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen3-4B": [
        {
          "name": "Qwen3-4B",
          "huggingface_id": "Qwen/Qwen3-4B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-4B-FP8": [
        {
          "name": "Qwen3-4B (fp8)",
          "huggingface_id": "Qwen/Qwen3-4B-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3-ASR-1.7B": [
        {
          "name": "Qwen3-ASR-1.7B",
          "huggingface_id": "Qwen/Qwen3-ASR-1.7B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (voicedesign-1.7b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "voicedesign-1.7b"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-1.7B-Base": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (base-1.7b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-1.7B-Base",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base-1.7b"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (customvoice-0.6b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "customvoice-0.6b"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-0.6B-Base": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (base-0.6b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-0.6B-Base",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base-0.6b"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-0.8B": [
        {
          "name": "Qwen3.5-0.8B",
          "huggingface_id": "Qwen/Qwen3.5-0.8B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-122B-A10B": [
        {
          "name": "Qwen3.5-122B-A10B",
          "huggingface_id": "Qwen/Qwen3.5-122B-A10B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-122B-A10B-FP8": [
        {
          "name": "Qwen3.5-122B-A10B (fp8)",
          "huggingface_id": "Qwen/Qwen3.5-122B-A10B-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3.5-122B-A10B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-122B-A10B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-122B-A10B-GPTQ-Int4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-27B": [
        {
          "name": "Qwen3.5-27B",
          "huggingface_id": "Qwen/Qwen3.5-27B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-27B-FP8": [
        {
          "name": "Qwen3.5-27B (fp8)",
          "huggingface_id": "Qwen/Qwen3.5-27B-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3.5-27B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-27B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-27B-GPTQ-Int4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-2B": [
        {
          "name": "Qwen3.5-2B",
          "huggingface_id": "Qwen/Qwen3.5-2B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-35B-A3B": [
        {
          "name": "Qwen3.5-35B-A3B",
          "huggingface_id": "Qwen/Qwen3.5-35B-A3B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-35B-A3B-FP8": [
        {
          "name": "Qwen3.5-35B-A3B (fp8)",
          "huggingface_id": "Qwen/Qwen3.5-35B-A3B-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3.5-35B-A3B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-35B-A3B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-35B-A3B-GPTQ-Int4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B": [
        {
          "name": "Qwen3.5-397B",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Qwen3.5-397B-A17B-NVFP4": [
        {
          "name": "Qwen3.5-397B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.5-397B-A17B-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-397B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-4B": [
        {
          "name": "Qwen3.5-4B",
          "huggingface_id": "Qwen/Qwen3.5-4B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-9B": [
        {
          "name": "Qwen3.5-9B",
          "huggingface_id": "Qwen/Qwen3.5-9B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-27B": [
        {
          "name": "Qwen3.6-27B",
          "huggingface_id": "Qwen/Qwen3.6-27B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-27B-FP8": [
        {
          "name": "Qwen3.6-27B (fp8)",
          "huggingface_id": "Qwen/Qwen3.6-27B-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "512",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3.6-27B-NVFP4": [
        {
          "name": "Qwen3.6-27B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.6-27B-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B": [
        {
          "name": "Qwen3.6-35B-A3B",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B-FP8": [
        {
          "name": "Qwen3.6-35B-A3B (fp8)",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3.6-35B-A3B-NVFP4": [
        {
          "name": "Qwen3.6-35B-A3B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.6-35B-A3B-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3Guard-Gen-8B": [
        {
          "name": "Qwen3Guard-Gen-8B",
          "huggingface_id": "Qwen/Qwen3Guard-Gen-8B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3Guard-Gen-4B": [
        {
          "name": "Qwen3Guard-Gen-8B (small_4b)",
          "huggingface_id": "Qwen/Qwen3Guard-Gen-4B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "small_4b"
        }
      ],
      "Qwen/Qwen3Guard-Gen-0.6B": [
        {
          "name": "Qwen3Guard-Gen-8B (tiny_0_6b)",
          "huggingface_id": "Qwen/Qwen3Guard-Gen-0.6B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "tiny_0_6b"
        }
      ],
      "Wan-AI/Wan2.2-T2V-A14B-Diffusers": [
        {
          "name": "Wan2.2",
          "huggingface_id": "Wan-AI/Wan2.2-T2V-A14B-Diffusers",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "XiaomiMiMo/MiMo-V2-Flash": [
        {
          "name": "MiMo-V2-Flash",
          "huggingface_id": "XiaomiMiMo/MiMo-V2-Flash",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "XiaomiMiMo/MiMo-V2.5-Pro": [
        {
          "name": "MiMo-V2.5-Pro",
          "huggingface_id": "XiaomiMiMo/MiMo-V2.5-Pro",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "XiaomiMiMo/MiMo-V2.5": [
        {
          "name": "MiMo-V2.5",
          "huggingface_id": "XiaomiMiMo/MiMo-V2.5",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "arcee-ai/Trinity-Large-Thinking": [
        {
          "name": "Trinity-Large-Thinking",
          "huggingface_id": "arcee-ai/Trinity-Large-Thinking",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "arcee-ai/Trinity-Large-Thinking-NVFP4": [
        {
          "name": "Trinity-Large-Thinking (nvfp4)",
          "huggingface_id": "arcee-ai/Trinity-Large-Thinking-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "baidu/ERNIE-4.5-21B-A3B-PT": [
        {
          "name": "ERNIE-4.5",
          "huggingface_id": "baidu/ERNIE-4.5-21B-A3B-PT",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "baidu/ERNIE-4.5-300B-A47B-PT": [
        {
          "name": "ERNIE-4.5 (300b)",
          "huggingface_id": "baidu/ERNIE-4.5-300B-A47B-PT",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--gpu-memory-utilization": "0.95",
                "--quantization": "fp8",
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--gpu-memory-utilization": "0.95",
                "--quantization": "fp8",
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "300b"
        }
      ],
      "baidu/ERNIE-4.5-VL-28B-A3B-PT": [
        {
          "name": "ERNIE-4.5-VL",
          "huggingface_id": "baidu/ERNIE-4.5-VL-28B-A3B-PT",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "baidu/ERNIE-4.5-VL-424B-A47B-PT": [
        {
          "name": "ERNIE-4.5-VL (424b)",
          "huggingface_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "424b"
        },
        {
          "name": "ERNIE-4.5-VL (424b_fp8)",
          "huggingface_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--quantization": "fp8",
                "--cpu-offload-gb": "50",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--quantization": "fp8",
                "--cpu-offload-gb": "50",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "424b_fp8"
        }
      ],
      "baidu/Unlimited-OCR": [
        {
          "name": "Unlimited-OCR",
          "huggingface_id": "baidu/Unlimited-OCR",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--logits_processors": "vllm.model_executor.models.unlimited_ocr:NGramPerReqLogitsProcessor",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "bosonai/higgs-audio-v3-tts-4b": [
        {
          "name": "Higgs Audio V3 TTS",
          "huggingface_id": "bosonai/higgs-audio-v3-tts-4b",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-OCR-2": [
        {
          "name": "DeepSeek-OCR-2",
          "huggingface_id": "deepseek-ai/DeepSeek-OCR-2",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--logits_processors": "vllm.model_executor.models.deepseek_ocr:NGramPerReqLogitsProcessor",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-OCR": [
        {
          "name": "DeepSeek-OCR",
          "huggingface_id": "deepseek-ai/DeepSeek-OCR",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--logits_processors": "vllm.model_executor.models.deepseek_ocr:NGramPerReqLogitsProcessor",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-R1": [
        {
          "name": "DeepSeek-R1",
          "huggingface_id": "deepseek-ai/DeepSeek-R1",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-R1-0528": [
        {
          "name": "DeepSeek-R1 (r1_0528)",
          "huggingface_id": "deepseek-ai/DeepSeek-R1-0528",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "r1_0528"
        }
      ],
      "nvidia/DeepSeek-R1-0528-NVFP4-v2": [
        {
          "name": "DeepSeek-R1 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-R1-0528-NVFP4-v2",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3.1": [
        {
          "name": "DeepSeek-V3.1",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.1",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3.1-NVFP4": [
        {
          "name": "DeepSeek-V3.1 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V3.1-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3.2-Exp": [
        {
          "name": "DeepSeek-V3.2-Exp",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.2-Exp",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-V3.2": [
        {
          "name": "DeepSeek-V3.2",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.2",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3.2-NVFP4": [
        {
          "name": "DeepSeek-V3.2 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V3.2-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--kv-cache-dtype": "fp8",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--kv-cache-dtype": "fp8",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3": [
        {
          "name": "DeepSeek-V3",
          "huggingface_id": "deepseek-ai/DeepSeek-V3",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3-FP4": [
        {
          "name": "DeepSeek-V3 (fp4)",
          "huggingface_id": "nvidia/DeepSeek-V3-FP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Flash": [
        {
          "name": "DeepSeek-V4-Flash",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Flash",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Flash-NVFP4": [
        {
          "name": "DeepSeek-V4-Flash (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Flash-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro": [
        {
          "name": "DeepSeek-V4-Pro",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Pro-NVFP4": [
        {
          "name": "DeepSeek-V4-Pro (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Pro-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro-DSpark": [
        {
          "name": "DeepSeek-V4-Pro (dspark)",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro-DSpark",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "dspark"
        }
      ],
      "fishaudio/s2-pro": [
        {
          "name": "Fish Speech S2 Pro",
          "huggingface_id": "fishaudio/s2-pro",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ling-2.6-1T": [
        {
          "name": "Ling-2.6-1T",
          "huggingface_id": "inclusionAI/Ling-2.6-1T",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ling-2.6-flash": [
        {
          "name": "Ling-2.6-flash",
          "huggingface_id": "inclusionAI/Ling-2.6-flash",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ming-omni-tts-0.5B": [
        {
          "name": "Ming-omni-tts-0.5B",
          "huggingface_id": "inclusionAI/Ming-omni-tts-0.5B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--enforce-eager": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ring-1T-FP8": [
        {
          "name": "Ring-1T-FP8",
          "huggingface_id": "inclusionAI/Ring-1T-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ring-2.6-1T": [
        {
          "name": "Ring-2.6-1T",
          "huggingface_id": "inclusionAI/Ring-2.6-1T",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S1": [
        {
          "name": "Intern-S1",
          "huggingface_id": "internlm/Intern-S1",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S1-FP8": [
        {
          "name": "Intern-S1 (fp8)",
          "huggingface_id": "internlm/Intern-S1-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "internlm/Intern-S2-Preview": [
        {
          "name": "Intern-S2-Preview",
          "huggingface_id": "internlm/Intern-S2-Preview",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S2-Preview-FP8": [
        {
          "name": "Intern-S2-Preview (fp8)",
          "huggingface_id": "internlm/Intern-S2-Preview-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-retrieval": [
        {
          "name": "Jina Embeddings v5 Text Small",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-retrieval",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-text-matching": [
        {
          "name": "Jina Embeddings v5 Text Small (text_matching)",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-text-matching",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "text_matching"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-classification": [
        {
          "name": "Jina Embeddings v5 Text Small (classification)",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-classification",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "classification"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-clustering": [
        {
          "name": "Jina Embeddings v5 Text Small (clustering)",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-clustering",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "clustering"
        }
      ],
      "jinaai/jina-reranker-m0": [
        {
          "name": "Jina Reranker m0",
          "huggingface_id": "jinaai/jina-reranker-m0",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--gpu-memory-utilization": "0.75",
                "--max-num-seqs": "32"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "meituan-longcat/LongCat-Image-Edit": [
        {
          "name": "LongCat-Image-Edit",
          "huggingface_id": "meituan-longcat/LongCat-Image-Edit",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "meta-llama/Llama-3.3-70B-Instruct": [
        {
          "name": "Llama-3.3-70B",
          "huggingface_id": "meta-llama/Llama-3.3-70B-Instruct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-FP8": [
        {
          "name": "Llama-3.3-70B (fp8)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-NVFP4": [
        {
          "name": "Llama-3.3-70B (nvfp4)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "meta-llama/Llama-4-Scout-17B-16E-Instruct": [
        {
          "name": "Llama-4-Scout",
          "huggingface_id": "meta-llama/Llama-4-Scout-17B-16E-Instruct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8": [
        {
          "name": "Llama-4-Scout (fp8)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4": [
        {
          "name": "Llama-4-Scout (nvfp4)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "microsoft/Phi-4-mini-instruct": [
        {
          "name": "Phi-4",
          "huggingface_id": "microsoft/Phi-4-mini-instruct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "microsoft/Phi-4-mini-reasoning": [
        {
          "name": "Phi-4 (mini_reasoning)",
          "huggingface_id": "microsoft/Phi-4-mini-reasoning",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mini_reasoning"
        }
      ],
      "microsoft/Phi-4-reasoning": [
        {
          "name": "Phi-4 (reasoning)",
          "huggingface_id": "microsoft/Phi-4-reasoning",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "reasoning"
        }
      ],
      "microsoft/Phi-4-multimodal-instruct": [
        {
          "name": "Phi-4 (multimodal)",
          "huggingface_id": "microsoft/Phi-4-multimodal-instruct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "multimodal"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Ministral-3-8B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (8b)",
          "huggingface_id": "mistralai/Ministral-3-8B-Instruct-2512",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "8b"
        }
      ],
      "mistralai/Ministral-3-3B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (3b)",
          "huggingface_id": "mistralai/Ministral-3-3B-Instruct-2512",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "3b"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512-FP8": [
        {
          "name": "Ministral-3-Instruct (fp8)",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "mistralai/Ministral-3-8B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning",
          "huggingface_id": "mistralai/Ministral-3-8B-Reasoning-2512",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Ministral-3-3B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning (3b)",
          "huggingface_id": "mistralai/Ministral-3-3B-Reasoning-2512",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "3b"
        }
      ],
      "mistralai/Ministral-3-14B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning (14b)",
          "huggingface_id": "mistralai/Ministral-3-14B-Reasoning-2512",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "14b"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512": [
        {
          "name": "Mistral-Large-3-675B-Instruct",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4": [
        {
          "name": "Mistral-Large-3-675B-Instruct (nvfp4)",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512-FP8": [
        {
          "name": "Mistral-Large-3-675B-Instruct (fp8)",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "mistralai/Mistral-Medium-3.5-128B": [
        {
          "name": "Mistral-Medium-3.5",
          "huggingface_id": "mistralai/Mistral-Medium-3.5-128B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Medium-3.5-128B-EAGLE\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Mistral-Small-4-119B-2603": [
        {
          "name": "Mistral-Small-4-119B",
          "huggingface_id": "mistralai/Mistral-Small-4-119B-2603",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "FLASH_ATTN_MLA",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "FLASH_ATTN_MLA",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Mistral-Small-4-119B-2603-NVFP4": [
        {
          "name": "Mistral-Small-4-119B (nvfp4)",
          "huggingface_id": "mistralai/Mistral-Small-4-119B-2603-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "TRITON_MLA",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "TRITON_MLA",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "mistralai/Voxtral-4B-TTS-2603": [
        {
          "name": "Voxtral-4B-TTS-2603",
          "huggingface_id": "mistralai/Voxtral-4B-TTS-2603",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Voxtral-Mini-4B-Realtime-2602": [
        {
          "name": "Voxtral-Mini-4B-Realtime-2602",
          "huggingface_id": "mistralai/Voxtral-Mini-4B-Realtime-2602",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer-mode": "mistral",
                "--compilation_config": "{\"cudagraph_mode\": \"PIECEWISE\"}"
              },
              "extra_envs": {
                "VLLM_DISABLE_COMPILE_CACHE": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-K2-Instruct": [
        {
          "name": "Kimi-K2-Instruct",
          "huggingface_id": "moonshotai/Kimi-K2-Instruct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-K2-Thinking": [
        {
          "name": "Kimi-K2-Thinking",
          "huggingface_id": "moonshotai/Kimi-K2-Thinking",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2-Thinking-NVFP4": [
        {
          "name": "Kimi-K2-Thinking (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2-Thinking-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.5": [
        {
          "name": "Kimi-K2.5",
          "huggingface_id": "moonshotai/Kimi-K2.5",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.5-NVFP4": [
        {
          "name": "Kimi-K2.5 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.5-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.6": [
        {
          "name": "Kimi-K2.6",
          "huggingface_id": "moonshotai/Kimi-K2.6",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.6-NVFP4": [
        {
          "name": "Kimi-K2.6 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.6-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.7-Code": [
        {
          "name": "Kimi-K2.7-Code",
          "huggingface_id": "moonshotai/Kimi-K2.7-Code",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-Linear-48B-A3B-Instruct": [
        {
          "name": "Kimi-Linear-48B-A3B-Instruct",
          "huggingface_id": "moonshotai/Kimi-Linear-48B-A3B-Instruct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Nano": [
        {
          "name": "Cosmos3-Nano",
          "huggingface_id": "nvidia/Cosmos3-Nano",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super-Image2Video": [
        {
          "name": "Cosmos3-Super-Image2Video",
          "huggingface_id": "nvidia/Cosmos3-Super-Image2Video",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super-Text2Image": [
        {
          "name": "Cosmos3-Super-Text2Image",
          "huggingface_id": "nvidia/Cosmos3-Super-Text2Image",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super": [
        {
          "name": "Cosmos3-Super",
          "huggingface_id": "nvidia/Cosmos3-Super",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Nano-30B-A3B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Nano-30B-A3B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_FLASHINFER_MOE_BACKEND": "throughput"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_FLASHINFER_MOE_BACKEND": "throughput"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Nano-4B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Nano-4B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (base_bf16)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base_bf16"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Ultra-550B-A55B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}"
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4": [
        {
          "name": "NVIDIA Nemotron-3-Ultra-550B-A55B (fp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}"
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-FP8": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-NVFP4-QAD": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-NVFP4-QAD",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-FP8": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-NVFP4": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Base": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (base)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Base",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Japanese": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (japanese)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Japanese",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "japanese"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning (fp8)",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning (nvfp4)",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "openai/gpt-oss-120b": [
        {
          "name": "GPT-OSS",
          "huggingface_id": "openai/gpt-oss-120b",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "amd/gpt-oss-120b-w-mxfp4-a-fp8": [
        {
          "name": "GPT-OSS (amd_fp8)",
          "huggingface_id": "amd/gpt-oss-120b-w-mxfp4-a-fp8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "amd_fp8"
        }
      ],
      "openai/gpt-oss-20b": [
        {
          "name": "GPT-OSS 20B",
          "huggingface_id": "openai/gpt-oss-20b",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "openbmb/MiniCPM-V-4.6": [
        {
          "name": "MiniCPM-V 4.6",
          "huggingface_id": "openbmb/MiniCPM-V-4.6",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "openbmb/MiniCPM5-1B": [
        {
          "name": "MiniCPM5-1B",
          "huggingface_id": "openbmb/MiniCPM5-1B",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "openbmb/VoxCPM2": [
        {
          "name": "VoxCPM2",
          "huggingface_id": "openbmb/VoxCPM2",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "pfnet/plamo-2-translate": [
        {
          "name": "PLaMo 2 Translate",
          "huggingface_id": "pfnet/plamo-2-translate",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "pfnet/plamo-3-nict-31b-base": [
        {
          "name": "PLaMo 3 NICT 31B Base",
          "huggingface_id": "pfnet/plamo-3-nict-31b-base",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-M.1": [
        {
          "name": "Laguna M.1",
          "huggingface_id": "poolside/Laguna-M.1",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-M.1-FP8": [
        {
          "name": "Laguna M.1 (fp8)",
          "huggingface_id": "poolside/Laguna-M.1-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "poolside/Laguna-M.1-NVFP4": [
        {
          "name": "Laguna M.1 (nvfp4)",
          "huggingface_id": "poolside/Laguna-M.1-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "poolside/Laguna-XS-2.1": [
        {
          "name": "Laguna XS-2.1",
          "huggingface_id": "poolside/Laguna-XS-2.1",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-XS-2.1-FP8": [
        {
          "name": "Laguna XS-2.1 (fp8)",
          "huggingface_id": "poolside/Laguna-XS-2.1-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "poolside/Laguna-XS-2.1-NVFP4": [
        {
          "name": "Laguna XS-2.1 (nvfp4)",
          "huggingface_id": "poolside/Laguna-XS-2.1-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "poolside/Laguna-XS-2.1-INT4": [
        {
          "name": "Laguna XS-2.1 (int4)",
          "huggingface_id": "poolside/Laguna-XS-2.1-INT4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "poolside/Laguna-XS.2": [
        {
          "name": "Laguna XS.2",
          "huggingface_id": "poolside/Laguna-XS.2",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-XS.2-FP8": [
        {
          "name": "Laguna XS.2 (fp8)",
          "huggingface_id": "poolside/Laguna-XS.2-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "poolside/Laguna-XS.2-NVFP4": [
        {
          "name": "Laguna XS.2 (nvfp4)",
          "huggingface_id": "poolside/Laguna-XS.2-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "poolside/Laguna-XS.2-INT4": [
        {
          "name": "Laguna XS.2 (int4)",
          "huggingface_id": "poolside/Laguna-XS.2-INT4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "stabilityai/stable-audio-open-1.0": [
        {
          "name": "Stable Audio Open",
          "huggingface_id": "stabilityai/stable-audio-open-1.0",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enforce-eager": "",
                "--gpu-memory-utilization": "0.9"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stabilityai/stable-diffusion-3.5-medium": [
        {
          "name": "Stable Diffusion 3.5",
          "huggingface_id": "stabilityai/stable-diffusion-3.5-medium",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stabilityai/stable-diffusion-3.5-large": [
        {
          "name": "Stable Diffusion 3.5 (large)",
          "huggingface_id": "stabilityai/stable-diffusion-3.5-large",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "large"
        }
      ],
      "stabilityai/stable-diffusion-3.5-large-turbo": [
        {
          "name": "Stable Diffusion 3.5 (large_turbo)",
          "huggingface_id": "stabilityai/stable-diffusion-3.5-large-turbo",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "large_turbo"
        }
      ],
      "stepfun-ai/Step-3.5-Flash": [
        {
          "name": "Step-3.5-Flash",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-FP8": [
        {
          "name": "Step-3.5-Flash (fp8)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-INT4": [
        {
          "name": "Step-3.5-Flash (int4)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-INT4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-INT8": [
        {
          "name": "Step-3.5-Flash (int8)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-INT8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int8"
        }
      ],
      "stepfun-ai/Step-3.7-Flash": [
        {
          "name": "Step-3.7-Flash",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stepfun-ai/Step-3.7-Flash-FP8": [
        {
          "name": "Step-3.7-Flash (fp8)",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "stepfun-ai/Step-3.7-Flash-NVFP4": [
        {
          "name": "Step-3.7-Flash (nvfp4)",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--quantization": "modelopt",
                "--kv-cache-dtype": "fp8",
                "--gpu-memory-utilization": "0.9",
                "--async-scheduling": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--quantization": "modelopt",
                "--kv-cache-dtype": "fp8",
                "--gpu-memory-utilization": "0.9",
                "--async-scheduling": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "tencent/Hunyuan-A13B-Instruct": [
        {
          "name": "Hunyuan-A13B-Instruct",
          "huggingface_id": "tencent/Hunyuan-A13B-Instruct",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hunyuan-A13B-Instruct-FP8": [
        {
          "name": "Hunyuan-A13B-Instruct (fp8)",
          "huggingface_id": "tencent/Hunyuan-A13B-Instruct-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "tencent/HunyuanOCR": [
        {
          "name": "HunyuanOCR",
          "huggingface_id": "tencent/HunyuanOCR",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3-preview": [
        {
          "name": "Hy3-preview",
          "huggingface_id": "tencent/Hy3-preview",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3": [
        {
          "name": "Hy3",
          "huggingface_id": "tencent/Hy3",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3-FP8": [
        {
          "name": "Hy3 (fp8)",
          "huggingface_id": "tencent/Hy3-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "thinkingmachines/Inkling-NVFP4": [
        {
          "name": "TML Inkling",
          "huggingface_id": "thinkingmachines/Inkling-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "thinkingmachines/Inkling": [
        {
          "name": "TML Inkling (bf16)",
          "huggingface_id": "thinkingmachines/Inkling",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "zai-org/GLM-4.5": [
        {
          "name": "GLM-4.5",
          "huggingface_id": "zai-org/GLM-4.5",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.5-FP8": [
        {
          "name": "GLM-4.5 (fp8)",
          "huggingface_id": "zai-org/GLM-4.5-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.5V": [
        {
          "name": "GLM-4.5V",
          "huggingface_id": "zai-org/GLM-4.5V",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.5V-FP8": [
        {
          "name": "GLM-4.5V (fp8)",
          "huggingface_id": "zai-org/GLM-4.5V-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.6": [
        {
          "name": "GLM-4.6",
          "huggingface_id": "zai-org/GLM-4.6",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.6-FP8": [
        {
          "name": "GLM-4.6 (fp8)",
          "huggingface_id": "zai-org/GLM-4.6-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.6V": [
        {
          "name": "GLM-4.6V",
          "huggingface_id": "zai-org/GLM-4.6V",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.6V-FP8": [
        {
          "name": "GLM-4.6V (fp8)",
          "huggingface_id": "zai-org/GLM-4.6V-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.7": [
        {
          "name": "GLM-4.7",
          "huggingface_id": "zai-org/GLM-4.7",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.7-FP8": [
        {
          "name": "GLM-4.7 (fp8)",
          "huggingface_id": "zai-org/GLM-4.7-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-4.7-NVFP4": [
        {
          "name": "GLM-4.7 (nvfp4)",
          "huggingface_id": "nvidia/GLM-4.7-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.1": [
        {
          "name": "GLM-5.1",
          "huggingface_id": "zai-org/GLM-5.1",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-5.1-FP8": [
        {
          "name": "GLM-5.1 (fp8)",
          "huggingface_id": "zai-org/GLM-5.1-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-5.1-NVFP4": [
        {
          "name": "GLM-5.1 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5.1-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.2-FP8": [
        {
          "name": "GLM-5.2",
          "huggingface_id": "zai-org/GLM-5.2-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/GLM-5.2-NVFP4": [
        {
          "name": "GLM-5.2 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5.2-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.2": [
        {
          "name": "GLM-5.2 (bf16)",
          "huggingface_id": "zai-org/GLM-5.2",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "zai-org/GLM-5": [
        {
          "name": "GLM-5",
          "huggingface_id": "zai-org/GLM-5",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-5-FP8": [
        {
          "name": "GLM-5 (fp8)",
          "huggingface_id": "zai-org/GLM-5-FP8",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-5-NVFP4": [
        {
          "name": "GLM-5 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-ASR-Nano-2512": [
        {
          "name": "GLM-ASR-Nano-2512",
          "huggingface_id": "zai-org/GLM-ASR-Nano-2512",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-GA": [
        {
          "name": "GLM-GA",
          "huggingface_id": "zai-org/GLM-GA",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--allowed-local-media-path": "/",
                "--mm-processor-cache-type": "shm",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "glm4_6v"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-Image": [
        {
          "name": "GLM-Image",
          "huggingface_id": "zai-org/GLM-Image",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-OCR": [
        {
          "name": "GLM-OCR",
          "huggingface_id": "zai-org/GLM-OCR",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-TTS": [
        {
          "name": "GLM-TTS",
          "huggingface_id": "zai-org/GLM-TTS",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/Glyph": [
        {
          "name": "Glyph",
          "huggingface_id": "zai-org/Glyph",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--limit-mm-per-prompt.video": "0",
                "--reasoning-parser": "glm45",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-26B-A4B-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-26B-A4B-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-31B-IT-NVFP4": [
        {
          "name": "Gemma 4 31B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-31B-IT-NVFP4",
          "accelerators": "intel",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ]
    },
    "nvidia": {
      "ByteDance-Seed/Seed-OSS-36B-Instruct": [
        {
          "name": "Seed-OSS-36B",
          "huggingface_id": "ByteDance-Seed/Seed-OSS-36B-Instruct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "seed_oss"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "google/diffusiongemma-26B-A4B-it": [
        {
          "name": "DiffusionGemma 26B-A4B IT",
          "huggingface_id": "google/diffusiongemma-26B-A4B-it",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/diffusiongemma-26B-A4B-it-FP8-dynamic": [
        {
          "name": "DiffusionGemma 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/diffusiongemma-26B-A4B-it-FP8-dynamic",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/diffusiongemma-26B-A4B-it-NVFP4": [
        {
          "name": "DiffusionGemma 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/diffusiongemma-26B-A4B-it-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--max-num-seqs": "4",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-12B-it": [
        {
          "name": "Gemma 4 12B IT",
          "huggingface_id": "google/gemma-4-12B-it",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-12B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "google/gemma-4-12B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 12B IT (w4a16)",
          "huggingface_id": "google/gemma-4-12B-it-qat-w4a16-ct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-12B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-26B-A4B-it": [
        {
          "name": "Gemma 4 26B-A4B IT",
          "huggingface_id": "google/gemma-4-26B-A4B-it",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 26B-A4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-FP8-dynamic",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "RedHatAI/gemma-4-26B-A4B-it-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (nvfp4)",
          "huggingface_id": "RedHatAI/gemma-4-26B-A4B-it-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it": [
        {
          "name": "Gemma 4 31B IT",
          "huggingface_id": "google/gemma-4-31B-it",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-31B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 31B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-31B-it-FP8-dynamic",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/gemma-4-31B-it-NVFP4": [
        {
          "name": "Gemma 4 31B IT (nvfp4)",
          "huggingface_id": "nvidia/gemma-4-31B-it-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "google/gemma-4-31B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 31B IT (w4a16)",
          "huggingface_id": "google/gemma-4-31B-it-qat-w4a16-ct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it": [
        {
          "name": "Gemma 4 E2B IT",
          "huggingface_id": "google/gemma-4-E2B-it",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E2B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E2B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E2B-it-FP8-dynamic",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E2B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E2B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-w4a16-ct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E2B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E2B IT (mobile)",
          "huggingface_id": "google/gemma-4-E2B-it-qat-mobile-ct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E2B-it-assistant\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "google/gemma-4-E4B-it": [
        {
          "name": "Gemma 4 E4B IT",
          "huggingface_id": "google/gemma-4-E4B-it",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/gemma-4-E4B-it-FP8-dynamic": [
        {
          "name": "Gemma 4 E4B IT (fp8)",
          "huggingface_id": "RedHatAI/gemma-4-E4B-it-FP8-dynamic",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "google/gemma-4-E4B-it-qat-w4a16-ct": [
        {
          "name": "Gemma 4 E4B IT (w4a16)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-w4a16-ct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "w4a16"
        }
      ],
      "google/gemma-4-E4B-it-qat-mobile-ct": [
        {
          "name": "Gemma 4 E4B IT (mobile)",
          "huggingface_id": "google/gemma-4-E4B-it-qat-mobile-ct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-E4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mobile"
        }
      ],
      "google/translategemma-27b-it": [
        {
          "name": "TranslateGemma 27B IT",
          "huggingface_id": "google/translategemma-27b-it",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Infomaniak-AI/vllm-translategemma-27b-it": [
        {
          "name": "TranslateGemma 27B IT (vllm_optimized)",
          "huggingface_id": "Infomaniak-AI/vllm-translategemma-27b-it",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "vllm_optimized"
        }
      ],
      "google/translategemma-4b-it": [
        {
          "name": "TranslateGemma 27B IT (small_4b)",
          "huggingface_id": "google/translategemma-4b-it",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "small_4b"
        }
      ],
      "JetBrains/Mellum2-12B-A2.5B-Instruct": [
        {
          "name": "Mellum2-12B-A2.5B-Instruct",
          "huggingface_id": "JetBrains/Mellum2-12B-A2.5B-Instruct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "JetBrains/Mellum2-12B-A2.5B-Thinking": [
        {
          "name": "Mellum2-12B-A2.5B-Thinking",
          "huggingface_id": "JetBrains/Mellum2-12B-A2.5B-Thinking",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Base": [
        {
          "name": "LFM2.5 1.2B Base",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Base",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Instruct": [
        {
          "name": "LFM2.5 1.2B Instruct",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Instruct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-JP-202606": [
        {
          "name": "LFM2.5 1.2B JP (202606)",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-JP-202606",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-JP": [
        {
          "name": "LFM2.5 1.2B JP",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-JP",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-1.2B-Thinking": [
        {
          "name": "LFM2.5 1.2B Thinking",
          "huggingface_id": "LiquidAI/LFM2.5-1.2B-Thinking",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-230M": [
        {
          "name": "LFM2.5 230M",
          "huggingface_id": "LiquidAI/LFM2.5-230M",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-350M": [
        {
          "name": "LFM2.5 350M",
          "huggingface_id": "LiquidAI/LFM2.5-350M",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-8B-A1B": [
        {
          "name": "LFM2.5 8B A1B",
          "huggingface_id": "LiquidAI/LFM2.5-8B-A1B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "lfm2",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-VL-1.6B": [
        {
          "name": "LFM2.5 VL 1.6B",
          "huggingface_id": "LiquidAI/LFM2.5-VL-1.6B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "LiquidAI/LFM2.5-VL-450M": [
        {
          "name": "LFM2.5 VL 450M",
          "huggingface_id": "LiquidAI/LFM2.5-VL-450M",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M2.1": [
        {
          "name": "MiniMax-M2.1",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.1",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M2.5": [
        {
          "name": "MiniMax-M2.5",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.5",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--compilation-config": "{\"cudagraph_mode\":\"PIECEWISE\"}",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--compilation-config": "{\"cudagraph_mode\":\"PIECEWISE\"}",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.5-NVFP4": [
        {
          "name": "MiniMax-M2.5 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.5-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--compilation-config": "{\"cudagraph_mode\":\"PIECEWISE\"}",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--compilation-config": "{\"cudagraph_mode\":\"PIECEWISE\"}",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2.7": [
        {
          "name": "MiniMax-M2.7",
          "huggingface_id": "MiniMaxAI/MiniMax-M2.7",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/MiniMax-M2.7-NVFP4": [
        {
          "name": "MiniMax-M2.7 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M2.7-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M2": [
        {
          "name": "MiniMax-M2",
          "huggingface_id": "MiniMaxAI/MiniMax-M2",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "RedHatAI/MiniMax-M2-NVFP4": [
        {
          "name": "MiniMax-M2 (nvfp4)",
          "huggingface_id": "RedHatAI/MiniMax-M2-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "minimax_m2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "MiniMaxAI/MiniMax-M3": [
        {
          "name": "MiniMax-M3",
          "huggingface_id": "MiniMaxAI/MiniMax-M3",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "MiniMaxAI/MiniMax-M3-MXFP8": [
        {
          "name": "MiniMax-M3 (mxfp8)",
          "huggingface_id": "MiniMaxAI/MiniMax-M3-MXFP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp8"
        }
      ],
      "nvidia/MiniMax-M3-NVFP4": [
        {
          "name": "MiniMax-M3 (nvfp4)",
          "huggingface_id": "nvidia/MiniMax-M3-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--trust-remote-code": "",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_FLOAT32_MATMUL_PRECISION": "high"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "amd/MiniMax-M3-MXFP4": [
        {
          "name": "MiniMax-M3 (mxfp4)",
          "huggingface_id": "amd/MiniMax-M3-MXFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--block-size": "128",
                "--tool-call-parser": "minimax_m3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "minimax_m3",
                "--speculative-config": "{\"method\": \"eagle3\", \"model\": \"Inferact/MiniMax-M3-EAGLE3\", \"num_speculative_tokens\": 3, \"attention_backend\": \"FLASH_ATTN\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mxfp4"
        }
      ],
      "OpenGVLab/InternVL3_5-8B": [
        {
          "name": "InternVL3.5",
          "huggingface_id": "OpenGVLab/InternVL3_5-8B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-SoundEffect": [
        {
          "name": "MOSS-SoundEffect",
          "huggingface_id": "OpenMOSS-Team/MOSS-SoundEffect",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_sound_effect.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-TTS-Realtime": [
        {
          "name": "MOSS-TTS-Realtime",
          "huggingface_id": "OpenMOSS-Team/MOSS-TTS-Realtime",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-TTS": [
        {
          "name": "MOSS-TTS",
          "huggingface_id": "OpenMOSS-Team/MOSS-TTS",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_tts.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-TTSD-v1.0": [
        {
          "name": "MOSS-TTSD-v1.0",
          "huggingface_id": "OpenMOSS-Team/MOSS-TTSD-v1.0",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_ttsd.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-Transcribe-Diarize": [
        {
          "name": "MOSS-Transcribe-Diarize",
          "huggingface_id": "OpenMOSS-Team/MOSS-Transcribe-Diarize",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "OpenMOSS-Team/MOSS-VoiceGenerator": [
        {
          "name": "MOSS-VoiceGenerator",
          "huggingface_id": "OpenMOSS-Team/MOSS-VoiceGenerator",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--deploy-config": "vllm_omni/deploy/moss_voice_generator.yaml"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "PaddlePaddle/PaddleOCR-VL-1.5": [
        {
          "name": "PaddleOCR-VL-1.5",
          "huggingface_id": "PaddlePaddle/PaddleOCR-VL-1.5",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-batched-tokens": "16384",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "PaddlePaddle/PaddleOCR-VL-1.6": [
        {
          "name": "PaddleOCR-VL-1.6",
          "huggingface_id": "PaddlePaddle/PaddleOCR-VL-1.6",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-batched-tokens": "16384",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "PaddlePaddle/PaddleOCR-VL": [
        {
          "name": "PaddleOCR-VL",
          "huggingface_id": "PaddlePaddle/PaddleOCR-VL",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-batched-tokens": "16384",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen-Image": [
        {
          "name": "Qwen-Image",
          "huggingface_id": "Qwen/Qwen-Image",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        },
        {
          "name": "Qwen-Image (fp8)",
          "huggingface_id": "Qwen/Qwen-Image",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "fp8",
                "--ignored-layers": "img_mlp"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        },
        {
          "name": "Qwen-Image (int8)",
          "huggingface_id": "Qwen/Qwen-Image",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "int8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int8"
        }
      ],
      "Qwen/Qwen2.5-VL-72B-Instruct": [
        {
          "name": "Qwen2.5-VL-72B-Instruct",
          "huggingface_id": "Qwen/Qwen2.5-VL-72B-Instruct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen2.5-VL-72B-Instruct-AWQ": [
        {
          "name": "Qwen2.5-VL-72B-Instruct (awq)",
          "huggingface_id": "Qwen/Qwen2.5-VL-72B-Instruct-AWQ",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "awq",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen2.5-VL-7B-Instruct": [
        {
          "name": "Qwen2.5-VL-7B-Instruct",
          "huggingface_id": "Qwen/Qwen2.5-VL-7B-Instruct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen2.5-VL-7B-Instruct-AWQ": [
        {
          "name": "Qwen2.5-VL-7B-Instruct (awq)",
          "huggingface_id": "Qwen/Qwen2.5-VL-7B-Instruct-AWQ",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "awq",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen3-235B-A22B-Instruct-2507": [
        {
          "name": "Qwen3-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-Instruct-2507",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-235B-A22B-FP8": [
        {
          "name": "Qwen3-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-235B-A22B-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4": [
        {
          "name": "Qwen3-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-235B-A22B-Instruct-2507-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-32B": [
        {
          "name": "Qwen3-32B",
          "huggingface_id": "Qwen/Qwen3-32B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-32B-FP8": [
        {
          "name": "Qwen3-32B (fp8)",
          "huggingface_id": "Qwen/Qwen3-32B-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3-32B-AWQ": [
        {
          "name": "Qwen3-32B (awq)",
          "huggingface_id": "Qwen/Qwen3-32B-AWQ",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--quantization": "awq",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "awq"
        }
      ],
      "Qwen/Qwen3-4B": [
        {
          "name": "Qwen3-4B",
          "huggingface_id": "Qwen/Qwen3-4B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-4B-FP8": [
        {
          "name": "Qwen3-4B (fp8)",
          "huggingface_id": "Qwen/Qwen3-4B-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3-ASR-1.7B": [
        {
          "name": "Qwen3-ASR-1.7B",
          "huggingface_id": "Qwen/Qwen3-ASR-1.7B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Coder-480B-A35B-Instruct-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Coder-480B-A35B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Coder-480B-A35B-Instruct-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-Next-80B-A3B-Instruct-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4": [
        {
          "name": "Qwen3-Next-80B-A3B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-Next-80B-A3B-Instruct-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "hermes",
                "--speculative-config": "{\"method\":\"qwen3_next_mtp\",\"num_speculative_tokens\":2}",
                "--no-enable-chunked-prefill": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (voicedesign-1.7b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "voicedesign-1.7b"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-1.7B-Base": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (base-1.7b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-1.7B-Base",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base-1.7b"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (customvoice-0.6b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "customvoice-0.6b"
        }
      ],
      "Qwen/Qwen3-TTS-12Hz-0.6B-Base": [
        {
          "name": "Qwen3-TTS-12Hz-1.7B-CustomVoice (base-0.6b)",
          "huggingface_id": "Qwen/Qwen3-TTS-12Hz-0.6B-Base",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base-0.6b"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (fp8)",
          "huggingface_id": "Qwen/Qwen3-VL-235B-A22B-Instruct-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4": [
        {
          "name": "Qwen3-VL-235B-A22B-Instruct (nvfp4)",
          "huggingface_id": "nvidia/Qwen3-VL-235B-A22B-Instruct-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-0.8B": [
        {
          "name": "Qwen3.5-0.8B",
          "huggingface_id": "Qwen/Qwen3.5-0.8B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-122B-A10B": [
        {
          "name": "Qwen3.5-122B-A10B",
          "huggingface_id": "Qwen/Qwen3.5-122B-A10B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-122B-A10B-FP8": [
        {
          "name": "Qwen3.5-122B-A10B (fp8)",
          "huggingface_id": "Qwen/Qwen3.5-122B-A10B-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3.5-122B-A10B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-122B-A10B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-122B-A10B-GPTQ-Int4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-27B": [
        {
          "name": "Qwen3.5-27B",
          "huggingface_id": "Qwen/Qwen3.5-27B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-27B-FP8": [
        {
          "name": "Qwen3.5-27B (fp8)",
          "huggingface_id": "Qwen/Qwen3.5-27B-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3.5-27B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-27B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-27B-GPTQ-Int4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-2B": [
        {
          "name": "Qwen3.5-2B",
          "huggingface_id": "Qwen/Qwen3.5-2B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-35B-A3B": [
        {
          "name": "Qwen3.5-35B-A3B",
          "huggingface_id": "Qwen/Qwen3.5-35B-A3B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-35B-A3B-FP8": [
        {
          "name": "Qwen3.5-35B-A3B (fp8)",
          "huggingface_id": "Qwen/Qwen3.5-35B-A3B-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "Qwen/Qwen3.5-35B-A3B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-35B-A3B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-35B-A3B-GPTQ-Int4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B": [
        {
          "name": "Qwen3.5-397B",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Qwen3.5-397B-A17B-NVFP4": [
        {
          "name": "Qwen3.5-397B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.5-397B-A17B-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency",
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4": [
        {
          "name": "Qwen3.5-397B (gptq_int4)",
          "huggingface_id": "Qwen/Qwen3.5-397B-A17B-GPTQ-Int4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_DEEP_GEMM_WARMUP": "skip",
                "VLLM_USE_DEEP_GEMM": "0",
                "VLLM_FLASHINFER_MOE_BACKEND": "latency"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "gptq_int4"
        }
      ],
      "Qwen/Qwen3.5-4B": [
        {
          "name": "Qwen3.5-4B",
          "huggingface_id": "Qwen/Qwen3.5-4B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.5-9B": [
        {
          "name": "Qwen3.5-9B",
          "huggingface_id": "Qwen/Qwen3.5-9B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-27B": [
        {
          "name": "Qwen3.6-27B",
          "huggingface_id": "Qwen/Qwen3.6-27B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-27B-FP8": [
        {
          "name": "Qwen3.6-27B (fp8)",
          "huggingface_id": "Qwen/Qwen3.6-27B-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "512",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3.6-27B-NVFP4": [
        {
          "name": "Qwen3.6-27B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.6-27B-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B": [
        {
          "name": "Qwen3.6-35B-A3B",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3.6-35B-A3B-FP8": [
        {
          "name": "Qwen3.6-35B-A3B (fp8)",
          "huggingface_id": "Qwen/Qwen3.6-35B-A3B-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Qwen3.6-35B-A3B-NVFP4": [
        {
          "name": "Qwen3.6-35B-A3B (nvfp4)",
          "huggingface_id": "nvidia/Qwen3.6-35B-A3B-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--attention-backend": "flashinfer",
                "--moe-backend": "marlin",
                "--enable-prefix-caching": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3,\"moe_backend\":\"triton\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "Qwen/Qwen3Guard-Gen-8B": [
        {
          "name": "Qwen3Guard-Gen-8B",
          "huggingface_id": "Qwen/Qwen3Guard-Gen-8B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "Qwen/Qwen3Guard-Gen-4B": [
        {
          "name": "Qwen3Guard-Gen-8B (small_4b)",
          "huggingface_id": "Qwen/Qwen3Guard-Gen-4B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "small_4b"
        }
      ],
      "Qwen/Qwen3Guard-Gen-0.6B": [
        {
          "name": "Qwen3Guard-Gen-8B (tiny_0_6b)",
          "huggingface_id": "Qwen/Qwen3Guard-Gen-0.6B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "tiny_0_6b"
        }
      ],
      "Wan-AI/Wan2.2-T2V-A14B-Diffusers": [
        {
          "name": "Wan2.2",
          "huggingface_id": "Wan-AI/Wan2.2-T2V-A14B-Diffusers",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "XiaomiMiMo/MiMo-V2-Flash": [
        {
          "name": "MiMo-V2-Flash",
          "huggingface_id": "XiaomiMiMo/MiMo-V2-Flash",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.9",
                "--tool-call-parser": "qwen3_xml",
                "--reasoning-parser": "qwen3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "XiaomiMiMo/MiMo-V2.5-Pro": [
        {
          "name": "MiMo-V2.5-Pro",
          "huggingface_id": "XiaomiMiMo/MiMo-V2.5-Pro",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "XiaomiMiMo/MiMo-V2.5": [
        {
          "name": "MiMo-V2.5",
          "huggingface_id": "XiaomiMiMo/MiMo-V2.5",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--generation-config": "vllm",
                "--gpu-memory-utilization": "0.95",
                "--tool-call-parser": "mimo",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "mimo",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "arcee-ai/Trinity-Large-Thinking": [
        {
          "name": "Trinity-Large-Thinking",
          "huggingface_id": "arcee-ai/Trinity-Large-Thinking",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "arcee-ai/Trinity-Large-Thinking-NVFP4": [
        {
          "name": "Trinity-Large-Thinking (nvfp4)",
          "huggingface_id": "arcee-ai/Trinity-Large-Thinking-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--dtype": "bfloat16",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "deepseek_r1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "baidu/ERNIE-4.5-21B-A3B-PT": [
        {
          "name": "ERNIE-4.5",
          "huggingface_id": "baidu/ERNIE-4.5-21B-A3B-PT",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "baidu/ERNIE-4.5-300B-A47B-PT": [
        {
          "name": "ERNIE-4.5 (300b)",
          "huggingface_id": "baidu/ERNIE-4.5-300B-A47B-PT",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--gpu-memory-utilization": "0.95",
                "--quantization": "fp8",
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--gpu-memory-utilization": "0.95",
                "--quantization": "fp8",
                "--speculative-config": "{\"method\":\"ernie_mtp\",\"model\":\"baidu/ERNIE-4.5-21B-A3B-PT\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "300b"
        }
      ],
      "baidu/ERNIE-4.5-VL-28B-A3B-PT": [
        {
          "name": "ERNIE-4.5-VL",
          "huggingface_id": "baidu/ERNIE-4.5-VL-28B-A3B-PT",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "baidu/ERNIE-4.5-VL-424B-A47B-PT": [
        {
          "name": "ERNIE-4.5-VL (424b)",
          "huggingface_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "424b"
        },
        {
          "name": "ERNIE-4.5-VL (424b_fp8)",
          "huggingface_id": "baidu/ERNIE-4.5-VL-424B-A47B-PT",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--quantization": "fp8",
                "--cpu-offload-gb": "50",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--quantization": "fp8",
                "--cpu-offload-gb": "50",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "424b_fp8"
        }
      ],
      "baidu/Unlimited-OCR": [
        {
          "name": "Unlimited-OCR",
          "huggingface_id": "baidu/Unlimited-OCR",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--logits_processors": "vllm.model_executor.models.unlimited_ocr:NGramPerReqLogitsProcessor",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "bosonai/higgs-audio-v3-tts-4b": [
        {
          "name": "Higgs Audio V3 TTS",
          "huggingface_id": "bosonai/higgs-audio-v3-tts-4b",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-OCR-2": [
        {
          "name": "DeepSeek-OCR-2",
          "huggingface_id": "deepseek-ai/DeepSeek-OCR-2",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--logits_processors": "vllm.model_executor.models.deepseek_ocr:NGramPerReqLogitsProcessor",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-OCR": [
        {
          "name": "DeepSeek-OCR",
          "huggingface_id": "deepseek-ai/DeepSeek-OCR",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--logits_processors": "vllm.model_executor.models.deepseek_ocr:NGramPerReqLogitsProcessor",
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-R1": [
        {
          "name": "DeepSeek-R1",
          "huggingface_id": "deepseek-ai/DeepSeek-R1",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-R1-0528": [
        {
          "name": "DeepSeek-R1 (r1_0528)",
          "huggingface_id": "deepseek-ai/DeepSeek-R1-0528",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "r1_0528"
        }
      ],
      "nvidia/DeepSeek-R1-0528-NVFP4-v2": [
        {
          "name": "DeepSeek-R1 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-R1-0528-NVFP4-v2",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekr1.jinja",
                "--reasoning-parser": "deepseek_r1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3.1": [
        {
          "name": "DeepSeek-V3.1",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.1",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3.1-NVFP4": [
        {
          "name": "DeepSeek-V3.1 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V3.1-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v31",
                "--chat-template": "examples/tool_chat_template_deepseekv31.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3.2-Exp": [
        {
          "name": "DeepSeek-V3.2-Exp",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.2-Exp",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "deepseek-ai/DeepSeek-V3.2": [
        {
          "name": "DeepSeek-V3.2",
          "huggingface_id": "deepseek-ai/DeepSeek-V3.2",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3.2-NVFP4": [
        {
          "name": "DeepSeek-V3.2 (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V3.2-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--kv-cache-dtype": "fp8",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--kv-cache-dtype": "fp8",
                "--tokenizer-mode": "deepseek_v32",
                "--tool-call-parser": "deepseek_v32",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V3": [
        {
          "name": "DeepSeek-V3",
          "huggingface_id": "deepseek-ai/DeepSeek-V3",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V3-FP4": [
        {
          "name": "DeepSeek-V3 (fp4)",
          "huggingface_id": "nvidia/DeepSeek-V3-FP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "deepseek_v3",
                "--chat-template": "examples/tool_chat_template_deepseekv3.jinja",
                "--reasoning-parser": "deepseek_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Flash": [
        {
          "name": "DeepSeek-V4-Flash",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Flash",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Flash-NVFP4": [
        {
          "name": "DeepSeek-V4-Flash (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Flash-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro": [
        {
          "name": "DeepSeek-V4-Pro",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/DeepSeek-V4-Pro-NVFP4": [
        {
          "name": "DeepSeek-V4-Pro (nvfp4)",
          "huggingface_id": "nvidia/DeepSeek-V4-Pro-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "deepseek-ai/DeepSeek-V4-Pro-DSpark": [
        {
          "name": "DeepSeek-V4-Pro (dspark)",
          "huggingface_id": "deepseek-ai/DeepSeek-V4-Pro-DSpark",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--block-size": "256",
                "--tokenizer-mode": "deepseek_v4",
                "--tool-call-parser": "deepseek_v4",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "deepseek_v4",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "dspark"
        }
      ],
      "fishaudio/s2-pro": [
        {
          "name": "Fish Speech S2 Pro",
          "huggingface_id": "fishaudio/s2-pro",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ling-2.6-1T": [
        {
          "name": "Ling-2.6-1T",
          "huggingface_id": "inclusionAI/Ling-2.6-1T",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ling-2.6-flash": [
        {
          "name": "Ling-2.6-flash",
          "huggingface_id": "inclusionAI/Ling-2.6-flash",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ming-omni-tts-0.5B": [
        {
          "name": "Ming-omni-tts-0.5B",
          "huggingface_id": "inclusionAI/Ming-omni-tts-0.5B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--enforce-eager": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ring-1T-FP8": [
        {
          "name": "Ring-1T-FP8",
          "huggingface_id": "inclusionAI/Ring-1T-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--max-num-seqs": "32",
                "--kv-cache-dtype": "fp8",
                "--served-model-name": "Ring-1T-FP8",
                "--gpu-memory-utilization": "0.97",
                "--compilation-config": "{\"use_inductor\": false}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "inclusionAI/Ring-2.6-1T": [
        {
          "name": "Ring-2.6-1T",
          "huggingface_id": "inclusionAI/Ring-2.6-1T",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S1": [
        {
          "name": "Intern-S1",
          "huggingface_id": "internlm/Intern-S1",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S1-FP8": [
        {
          "name": "Intern-S1 (fp8)",
          "huggingface_id": "internlm/Intern-S1-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "internlm",
                "--reasoning-parser": "deepseek_r1",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "internlm/Intern-S2-Preview": [
        {
          "name": "Intern-S2-Preview",
          "huggingface_id": "internlm/Intern-S2-Preview",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "internlm/Intern-S2-Preview-FP8": [
        {
          "name": "Intern-S2-Preview (fp8)",
          "huggingface_id": "internlm/Intern-S2-Preview-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "qwen3",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-retrieval": [
        {
          "name": "Jina Embeddings v5 Text Small",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-retrieval",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-text-matching": [
        {
          "name": "Jina Embeddings v5 Text Small (text_matching)",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-text-matching",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "text_matching"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-classification": [
        {
          "name": "Jina Embeddings v5 Text Small (classification)",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-classification",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "classification"
        }
      ],
      "jinaai/jina-embeddings-v5-text-small-clustering": [
        {
          "name": "Jina Embeddings v5 Text Small (clustering)",
          "huggingface_id": "jinaai/jina-embeddings-v5-text-small-clustering",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--runner": "pooling"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "clustering"
        }
      ],
      "jinaai/jina-reranker-m0": [
        {
          "name": "Jina Reranker m0",
          "huggingface_id": "jinaai/jina-reranker-m0",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--gpu-memory-utilization": "0.75",
                "--max-num-seqs": "32"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "meituan-longcat/LongCat-Image-Edit": [
        {
          "name": "LongCat-Image-Edit",
          "huggingface_id": "meituan-longcat/LongCat-Image-Edit",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "meta-llama/Llama-3.3-70B-Instruct": [
        {
          "name": "Llama-3.3-70B",
          "huggingface_id": "meta-llama/Llama-3.3-70B-Instruct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-FP8": [
        {
          "name": "Llama-3.3-70B (fp8)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-3.3-70B-Instruct-NVFP4": [
        {
          "name": "Llama-3.3-70B (nvfp4)",
          "huggingface_id": "nvidia/Llama-3.3-70B-Instruct-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "meta-llama/Llama-4-Scout-17B-16E-Instruct": [
        {
          "name": "Llama-4-Scout",
          "huggingface_id": "meta-llama/Llama-4-Scout-17B-16E-Instruct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8": [
        {
          "name": "Llama-4-Scout (fp8)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4": [
        {
          "name": "Llama-4-Scout (nvfp4)",
          "huggingface_id": "nvidia/Llama-4-Scout-17B-16E-Instruct-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "microsoft/Phi-4-mini-instruct": [
        {
          "name": "Phi-4",
          "huggingface_id": "microsoft/Phi-4-mini-instruct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "microsoft/Phi-4-mini-reasoning": [
        {
          "name": "Phi-4 (mini_reasoning)",
          "huggingface_id": "microsoft/Phi-4-mini-reasoning",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "mini_reasoning"
        }
      ],
      "microsoft/Phi-4-reasoning": [
        {
          "name": "Phi-4 (reasoning)",
          "huggingface_id": "microsoft/Phi-4-reasoning",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "reasoning"
        }
      ],
      "microsoft/Phi-4-multimodal-instruct": [
        {
          "name": "Phi-4 (multimodal)",
          "huggingface_id": "microsoft/Phi-4-multimodal-instruct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "multimodal"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Ministral-3-8B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (8b)",
          "huggingface_id": "mistralai/Ministral-3-8B-Instruct-2512",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "8b"
        }
      ],
      "mistralai/Ministral-3-3B-Instruct-2512": [
        {
          "name": "Ministral-3-Instruct (3b)",
          "huggingface_id": "mistralai/Ministral-3-3B-Instruct-2512",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "3b"
        }
      ],
      "mistralai/Ministral-3-14B-Instruct-2512-FP8": [
        {
          "name": "Ministral-3-Instruct (fp8)",
          "huggingface_id": "mistralai/Ministral-3-14B-Instruct-2512-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "mistralai/Ministral-3-8B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning",
          "huggingface_id": "mistralai/Ministral-3-8B-Reasoning-2512",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Ministral-3-3B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning (3b)",
          "huggingface_id": "mistralai/Ministral-3-3B-Reasoning-2512",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "3b"
        }
      ],
      "mistralai/Ministral-3-14B-Reasoning-2512": [
        {
          "name": "Ministral-3-Reasoning (14b)",
          "huggingface_id": "mistralai/Ministral-3-14B-Reasoning-2512",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "14b"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512": [
        {
          "name": "Mistral-Large-3-675B-Instruct",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4": [
        {
          "name": "Mistral-Large-3-675B-Instruct (nvfp4)",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "mistralai/Mistral-Large-3-675B-Instruct-2512-FP8": [
        {
          "name": "Mistral-Large-3-675B-Instruct (fp8)",
          "huggingface_id": "mistralai/Mistral-Large-3-675B-Instruct-2512-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "mistralai/Mistral-Medium-3.5-128B": [
        {
          "name": "Mistral-Medium-3.5",
          "huggingface_id": "mistralai/Mistral-Medium-3.5-128B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer_mode": "mistral",
                "--config_format": "mistral",
                "--load_format": "mistral",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Medium-3.5-128B-EAGLE\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Mistral-Small-4-119B-2603": [
        {
          "name": "Mistral-Small-4-119B",
          "huggingface_id": "mistralai/Mistral-Small-4-119B-2603",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "FLASH_ATTN_MLA",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "FLASH_ATTN_MLA",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Mistral-Small-4-119B-2603-NVFP4": [
        {
          "name": "Mistral-Small-4-119B (nvfp4)",
          "huggingface_id": "mistralai/Mistral-Small-4-119B-2603-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--attention-backend": "TRITON_MLA",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--attention-backend": "TRITON_MLA",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "mistral",
                "--reasoning-parser": "mistral",
                "--speculative-config": "{\"model\":\"mistralai/Mistral-Small-4-119B-2603-eagle\",\"num_speculative_tokens\":3,\"method\":\"eagle\",\"max_model_len\":\"65536\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "mistralai/Voxtral-4B-TTS-2603": [
        {
          "name": "Voxtral-4B-TTS-2603",
          "huggingface_id": "mistralai/Voxtral-4B-TTS-2603",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "mistralai/Voxtral-Mini-4B-Realtime-2602": [
        {
          "name": "Voxtral-Mini-4B-Realtime-2602",
          "huggingface_id": "mistralai/Voxtral-Mini-4B-Realtime-2602",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tokenizer-mode": "mistral",
                "--compilation_config": "{\"cudagraph_mode\": \"PIECEWISE\"}"
              },
              "extra_envs": {
                "VLLM_DISABLE_COMPILE_CACHE": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-K2-Instruct": [
        {
          "name": "Kimi-K2-Instruct",
          "huggingface_id": "moonshotai/Kimi-K2-Instruct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-K2-Thinking": [
        {
          "name": "Kimi-K2-Thinking",
          "huggingface_id": "moonshotai/Kimi-K2-Thinking",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2-Thinking-NVFP4": [
        {
          "name": "Kimi-K2-Thinking (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2-Thinking-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "kimi_k2",
                "--reasoning-parser": "kimi_k2",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.5": [
        {
          "name": "Kimi-K2.5",
          "huggingface_id": "moonshotai/Kimi-K2.5",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.5-NVFP4": [
        {
          "name": "Kimi-K2.5 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.5-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.5-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.6": [
        {
          "name": "Kimi-K2.6",
          "huggingface_id": "moonshotai/Kimi-K2.6",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Kimi-K2.6-NVFP4": [
        {
          "name": "Kimi-K2.6 (nvfp4)",
          "huggingface_id": "nvidia/Kimi-K2.6-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.6-eagle3-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "moonshotai/Kimi-K2.7-Code": [
        {
          "name": "Kimi-K2.7-Code",
          "huggingface_id": "moonshotai/Kimi-K2.7-Code",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "kimi_k2",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "kimi_k2",
                "--mm-encoder-tp-mode": "data",
                "--speculative-config": "{\"model\":\"lightseekorg/kimi-k2.7-coder-eagle3.1-mla\",\"method\":\"eagle3\",\"num_speculative_tokens\":3}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "moonshotai/Kimi-Linear-48B-A3B-Instruct": [
        {
          "name": "Kimi-Linear-48B-A3B-Instruct",
          "huggingface_id": "moonshotai/Kimi-Linear-48B-A3B-Instruct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Nano": [
        {
          "name": "Cosmos3-Nano",
          "huggingface_id": "nvidia/Cosmos3-Nano",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super-Image2Video": [
        {
          "name": "Cosmos3-Super-Image2Video",
          "huggingface_id": "nvidia/Cosmos3-Super-Image2Video",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super-Text2Image": [
        {
          "name": "Cosmos3-Super-Text2Image",
          "huggingface_id": "nvidia/Cosmos3-Super-Text2Image",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Cosmos3-Super": [
        {
          "name": "Cosmos3-Super",
          "huggingface_id": "nvidia/Cosmos3-Super",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--host": "0.0.0.0",
                "--port": "8000",
                "--cfg-parallel-size": "2",
                "--ulysses-degree": "4",
                "--use-hsdp": "",
                "--hsdp-shard-size": "8",
                "--init-timeout": "1800"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Nano-30B-A3B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Nano-30B-A3B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_FLASHINFER_MOE_BACKEND": "throughput"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP8": "1",
                "VLLM_FLASHINFER_MOE_BACKEND": "throughput"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Nano-4B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "auto",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Nano-4B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--async-scheduling": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser-plugin": "nano_v3_reasoning_parser.py",
                "--reasoning-parser": "nano_v3"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-BF16",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Super-120B-A12B (base_bf16)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-Base-BF16",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base_bf16"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16": [
        {
          "name": "NVIDIA Nemotron-3-Ultra-550B-A55B",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}"
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4": [
        {
          "name": "NVIDIA Nemotron-3-Ultra-550B-A55B (fp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}"
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--served-model-name": "nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B",
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--max-num-seqs": "16",
                "--gpu-memory-utilization": "0.90",
                "--max-num-batched-tokens": "32768",
                "--enable-flashinfer-autotune": "",
                "--async-scheduling": "",
                "--mamba-backend": "triton",
                "--mamba-ssm-cache-dtype": "float32",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":5}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-BF16",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-FP8": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-NVFP4-QAD": [
        {
          "name": "NVIDIA Nemotron-Nano-12B-v2-VL (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-12B-v2-VL-NVFP4-QAD",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "opencv"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-FP8": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (fp8)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-NVFP4": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (nvfp4)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Base": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (base)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Base",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "base"
        }
      ],
      "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Japanese": [
        {
          "name": "NVIDIA Nemotron-Nano-9B-v2 (japanese)",
          "huggingface_id": "nvidia/NVIDIA-Nemotron-Nano-9B-v2-Japanese",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser-plugin": "nemotron_toolcall_parser_no_streaming.py",
                "--tool-call-parser": "nemotron_json"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "japanese"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning (fp8)",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4": [
        {
          "name": "Nemotron-3-Nano-Omni-30B-A3B-Reasoning (nvfp4)",
          "huggingface_id": "nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--media-io-kwargs": "{\"video\": {\"num_frames\": 512, \"fps\": 1}}",
                "--video-pruning-rate": "0.5",
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder",
                "--reasoning-parser": "nemotron_v3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "openai/gpt-oss-120b": [
        {
          "name": "GPT-OSS",
          "huggingface_id": "openai/gpt-oss-120b",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "amd/gpt-oss-120b-w-mxfp4-a-fp8": [
        {
          "name": "GPT-OSS (amd_fp8)",
          "huggingface_id": "amd/gpt-oss-120b-w-mxfp4-a-fp8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--speculative-config": "{\"model\":\"nvidia/gpt-oss-120b-Eagle3-v3\",\"num_speculative_tokens\":7,\"method\":\"eagle3\",\"draft_tensor_parallel_size\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "amd_fp8"
        }
      ],
      "openai/gpt-oss-20b": [
        {
          "name": "GPT-OSS 20B",
          "huggingface_id": "openai/gpt-oss-20b",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "openai",
                "--enable-auto-tool-choice": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "openbmb/MiniCPM-V-4.6": [
        {
          "name": "MiniCPM-V 4.6",
          "huggingface_id": "openbmb/MiniCPM-V-4.6",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "qwen3_coder"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "openbmb/MiniCPM5-1B": [
        {
          "name": "MiniCPM5-1B",
          "huggingface_id": "openbmb/MiniCPM5-1B",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "openbmb/VoxCPM2": [
        {
          "name": "VoxCPM2",
          "huggingface_id": "openbmb/VoxCPM2",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "pfnet/plamo-2-translate": [
        {
          "name": "PLaMo 2 Translate",
          "huggingface_id": "pfnet/plamo-2-translate",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "pfnet/plamo-3-nict-31b-base": [
        {
          "name": "PLaMo 3 NICT 31B Base",
          "huggingface_id": "pfnet/plamo-3-nict-31b-base",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-M.1": [
        {
          "name": "Laguna M.1",
          "huggingface_id": "poolside/Laguna-M.1",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-M.1-FP8": [
        {
          "name": "Laguna M.1 (fp8)",
          "huggingface_id": "poolside/Laguna-M.1-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "poolside/Laguna-M.1-NVFP4": [
        {
          "name": "Laguna M.1 (nvfp4)",
          "huggingface_id": "poolside/Laguna-M.1-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--default-chat-template-kwargs": "{\"enable_thinking\": true}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "poolside/Laguna-XS-2.1": [
        {
          "name": "Laguna XS-2.1",
          "huggingface_id": "poolside/Laguna-XS-2.1",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-XS-2.1-FP8": [
        {
          "name": "Laguna XS-2.1 (fp8)",
          "huggingface_id": "poolside/Laguna-XS-2.1-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "poolside/Laguna-XS-2.1-NVFP4": [
        {
          "name": "Laguna XS-2.1 (nvfp4)",
          "huggingface_id": "poolside/Laguna-XS-2.1-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "poolside/Laguna-XS-2.1-INT4": [
        {
          "name": "Laguna XS-2.1 (int4)",
          "huggingface_id": "poolside/Laguna-XS-2.1-INT4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS-2.1-DFlash\",\"num_speculative_tokens\":15,\"method\":\"dflash\"}",
                "--moe-backend": "triton",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "poolside/Laguna-XS.2": [
        {
          "name": "Laguna XS.2",
          "huggingface_id": "poolside/Laguna-XS.2",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "poolside/Laguna-XS.2-FP8": [
        {
          "name": "Laguna XS.2 (fp8)",
          "huggingface_id": "poolside/Laguna-XS.2-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_BLOCKSCALE_FP8_GEMM_FLASHINFER": "0"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "poolside/Laguna-XS.2-NVFP4": [
        {
          "name": "Laguna XS.2 (nvfp4)",
          "huggingface_id": "poolside/Laguna-XS.2-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "poolside/Laguna-XS.2-INT4": [
        {
          "name": "Laguna XS.2 (int4)",
          "huggingface_id": "poolside/Laguna-XS.2-INT4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "poolside_v1",
                "--reasoning-parser": "poolside_v1",
                "--speculative-config": "{\"model\":\"poolside/Laguna-XS.2-speculator.dflash\",\"num_speculative_tokens\":7,\"method\":\"dflash\"}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "stabilityai/stable-audio-open-1.0": [
        {
          "name": "Stable Audio Open",
          "huggingface_id": "stabilityai/stable-audio-open-1.0",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enforce-eager": "",
                "--gpu-memory-utilization": "0.9"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stabilityai/stable-diffusion-3.5-medium": [
        {
          "name": "Stable Diffusion 3.5",
          "huggingface_id": "stabilityai/stable-diffusion-3.5-medium",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stabilityai/stable-diffusion-3.5-large": [
        {
          "name": "Stable Diffusion 3.5 (large)",
          "huggingface_id": "stabilityai/stable-diffusion-3.5-large",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "large"
        }
      ],
      "stabilityai/stable-diffusion-3.5-large-turbo": [
        {
          "name": "Stable Diffusion 3.5 (large_turbo)",
          "huggingface_id": "stabilityai/stable-diffusion-3.5-large-turbo",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "large_turbo"
        }
      ],
      "stepfun-ai/Step-3.5-Flash": [
        {
          "name": "Step-3.5-Flash",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-FP8": [
        {
          "name": "Step-3.5-Flash (fp8)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-INT4": [
        {
          "name": "Step-3.5-Flash (int4)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-INT4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int4"
        }
      ],
      "stepfun-ai/Step-3.5-Flash-INT8": [
        {
          "name": "Step-3.5-Flash (int8)",
          "huggingface_id": "stepfun-ai/Step-3.5-Flash-INT8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--hf-overrides": "{\"num_nextn_predict_layers\": 1}",
                "--speculative-config": "{\"method\": \"step3p5_mtp\", \"num_speculative_tokens\": 1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "int8"
        }
      ],
      "stepfun-ai/Step-3.7-Flash": [
        {
          "name": "Step-3.7-Flash",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "stepfun-ai/Step-3.7-Flash-FP8": [
        {
          "name": "Step-3.7-Flash (fp8)",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "stepfun-ai/Step-3.7-Flash-NVFP4": [
        {
          "name": "Step-3.7-Flash (nvfp4)",
          "huggingface_id": "stepfun-ai/Step-3.7-Flash-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--quantization": "modelopt",
                "--kv-cache-dtype": "fp8",
                "--gpu-memory-utilization": "0.9",
                "--async-scheduling": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--disable-cascade-attn": "",
                "--quantization": "modelopt",
                "--kv-cache-dtype": "fp8",
                "--gpu-memory-utilization": "0.9",
                "--async-scheduling": "",
                "--tool-call-parser": "step3p5",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "step3p5",
                "--speculative-config": "{\"method\": \"mtp\", \"num_speculative_tokens\": 3}"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "tencent/Hunyuan-A13B-Instruct": [
        {
          "name": "Hunyuan-A13B-Instruct",
          "huggingface_id": "tencent/Hunyuan-A13B-Instruct",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hunyuan-A13B-Instruct-FP8": [
        {
          "name": "Hunyuan-A13B-Instruct (fp8)",
          "huggingface_id": "tencent/Hunyuan-A13B-Instruct-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "tencent/HunyuanOCR": [
        {
          "name": "HunyuanOCR",
          "huggingface_id": "tencent/HunyuanOCR",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3-preview": [
        {
          "name": "Hy3-preview",
          "huggingface_id": "tencent/Hy3-preview",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":1}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3": [
        {
          "name": "Hy3",
          "huggingface_id": "tencent/Hy3",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "tencent/Hy3-FP8": [
        {
          "name": "Hy3 (fp8)",
          "huggingface_id": "tencent/Hy3-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--tool-call-parser": "hy_v3",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "hy_v3",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":2}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "thinkingmachines/Inkling-NVFP4": [
        {
          "name": "TML Inkling",
          "huggingface_id": "thinkingmachines/Inkling-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "thinkingmachines/Inkling": [
        {
          "name": "TML Inkling (bf16)",
          "huggingface_id": "thinkingmachines/Inkling",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tokenizer-mode": "inkling",
                "--kernel-config.enable_flashinfer_autotune=False": "",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "inkling",
                "--reasoning-parser": "inkling",
                "--speculative-config": "{\"method\":\"mtp\",\"num_speculative_tokens\":8}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1",
                "FLASH_ATTENTION_CUTE_DSL_CACHE_ENABLED": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "zai-org/GLM-4.5": [
        {
          "name": "GLM-4.5",
          "huggingface_id": "zai-org/GLM-4.5",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.5-FP8": [
        {
          "name": "GLM-4.5 (fp8)",
          "huggingface_id": "zai-org/GLM-4.5-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.5V": [
        {
          "name": "GLM-4.5V",
          "huggingface_id": "zai-org/GLM-4.5V",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.5V-FP8": [
        {
          "name": "GLM-4.5V (fp8)",
          "huggingface_id": "zai-org/GLM-4.5V-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.6": [
        {
          "name": "GLM-4.6",
          "huggingface_id": "zai-org/GLM-4.6",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.6-FP8": [
        {
          "name": "GLM-4.6 (fp8)",
          "huggingface_id": "zai-org/GLM-4.6-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.6V": [
        {
          "name": "GLM-4.6V",
          "huggingface_id": "zai-org/GLM-4.6V",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.6V-FP8": [
        {
          "name": "GLM-4.6V (fp8)",
          "huggingface_id": "zai-org/GLM-4.6V-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--enable-expert-parallel": "",
                "--allowed-local-media-path": "/",
                "--mm-encoder-tp-mode": "data",
                "--mm-processor-cache-type": "shm",
                "--tool-call-parser": "glm45",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "zai-org/GLM-4.7": [
        {
          "name": "GLM-4.7",
          "huggingface_id": "zai-org/GLM-4.7",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-4.7-FP8": [
        {
          "name": "GLM-4.7 (fp8)",
          "huggingface_id": "zai-org/GLM-4.7-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-4.7-NVFP4": [
        {
          "name": "GLM-4.7 (nvfp4)",
          "huggingface_id": "nvidia/GLM-4.7-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.1": [
        {
          "name": "GLM-5.1",
          "huggingface_id": "zai-org/GLM-5.1",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-5.1-FP8": [
        {
          "name": "GLM-5.1 (fp8)",
          "huggingface_id": "zai-org/GLM-5.1-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-5.1-NVFP4": [
        {
          "name": "GLM-5.1 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5.1-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.2-FP8": [
        {
          "name": "GLM-5.2",
          "huggingface_id": "zai-org/GLM-5.2-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/GLM-5.2-NVFP4": [
        {
          "name": "GLM-5.2 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5.2-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-5.2": [
        {
          "name": "GLM-5.2 (bf16)",
          "huggingface_id": "zai-org/GLM-5.2",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8_e4m3",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "5",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "bf16"
        }
      ],
      "zai-org/GLM-5": [
        {
          "name": "GLM-5",
          "huggingface_id": "zai-org/GLM-5",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-5-FP8": [
        {
          "name": "GLM-5 (fp8)",
          "huggingface_id": "zai-org/GLM-5-FP8",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {}
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "fp8"
        }
      ],
      "nvidia/GLM-5-NVFP4": [
        {
          "name": "GLM-5 (nvfp4)",
          "huggingface_id": "nvidia/GLM-5-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3"
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--trust-remote-code": "",
                "--chat-template-content-format=string": "",
                "--kv-cache-dtype": "fp8",
                "--tool-call-parser": "glm47",
                "--enable-auto-tool-choice": "",
                "--reasoning-parser": "glm45",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "3",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_FLASHINFER_MOE_FP4": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "nvfp4"
        }
      ],
      "zai-org/GLM-ASR-Nano-2512": [
        {
          "name": "GLM-ASR-Nano-2512",
          "huggingface_id": "zai-org/GLM-ASR-Nano-2512",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {},
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-GA": [
        {
          "name": "GLM-GA",
          "huggingface_id": "zai-org/GLM-GA",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": "",
                "--allowed-local-media-path": "/",
                "--mm-processor-cache-type": "shm",
                "--reasoning-parser": "glm45"
              },
              "extra_envs": {
                "VLLM_VIDEO_LOADER_BACKEND": "glm4_6v"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-Image": [
        {
          "name": "GLM-Image",
          "huggingface_id": "zai-org/GLM-Image",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-OCR": [
        {
          "name": "GLM-OCR",
          "huggingface_id": "zai-org/GLM-OCR",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--mm-encoder-tp-mode": "data",
                "--speculative-config.method": "mtp",
                "--speculative-config.num_speculative_tokens": "1"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/GLM-TTS": [
        {
          "name": "GLM-TTS",
          "huggingface_id": "zai-org/GLM-TTS",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--omni": "",
                "--trust-remote-code": ""
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "zai-org/Glyph": [
        {
          "name": "Glyph",
          "huggingface_id": "zai-org/Glyph",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--no-enable-prefix-caching": "",
                "--mm-processor-cache-gb": "0",
                "--limit-mm-per-prompt.video": "0",
                "--reasoning-parser": "glm45",
                "--mm-encoder-tp-mode": "data"
              },
              "extra_envs": {}
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-26B-A4B-NVFP4": [
        {
          "name": "Gemma 4 26B-A4B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-26B-A4B-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            },
            "balance": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-26B-A4B-it-assistant\",\"num_speculative_tokens\":4}",
                "--enable-expert-parallel": ""
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ],
      "nvidia/Gemma-4-31B-IT-NVFP4": [
        {
          "name": "Gemma 4 31B IT (NVFP4)",
          "huggingface_id": "nvidia/Gemma-4-31B-IT-NVFP4",
          "accelerators": "nvidia",
          "provider": "vllm",
          "custom_image": "vllm/vllm-openai:v0.25.1",
          "strategies": {
            "latency-focus": {
              "extra_args": {
                "--kv-cache-dtype": "fp8",
                "--enable-auto-tool-choice": "",
                "--tool-call-parser": "gemma4",
                "--chat-template": "/vllm-workspace/examples/tool_chat_template_gemma4.jinja",
                "--reasoning-parser": "gemma4",
                "--speculative-config": "{\"model\":\"google/gemma-4-31B-it-assistant\",\"num_speculative_tokens\":4}"
              },
              "extra_envs": {
                "VLLM_USE_V2_MODEL_RUNNER": "1"
              }
            }
          },
          "default_strategy": "latency-focus",
          "variant_name": "default"
        }
      ]
    }
  }
}