{
  "_comment": "Native multi-LLM harness registry. Costs are USD per 1,000,000 tokens (in/out). Edit freely. Opus 4.8 reference: ~15/75 per 1M. Trust is execution-only — NONE of these are for orchestration or whole-project context. Pricing last verified from official provider pages 2026-06-01; run `scrooge-drift` to detect when model ids drift.",
  "version": 1,
  "providers": {
    "deepseek": {
      "env": [
        "DEEPSEEK_API_KEY"
      ],
      "base_url": "https://api.deepseek.com/v1",
      "openai_compatible": true
    },
    "kimi": {
      "env": [
        "MOONSHOT_API_KEY",
        "KIMI_API_KEY"
      ],
      "base_url": "https://api.moonshot.ai/v1",
      "openai_compatible": true
    },
    "zai": {
      "env": [
        "ZAI_API_KEY"
      ],
      "base_url": "https://api.z.ai/api/paas/v4",
      "openai_compatible": true
    },
    "gemini": {
      "env": [
        "GOOGLE_API_KEY",
        "GEMINI_API_KEY"
      ],
      "base_url": "https://generativelanguage.googleapis.com/v1beta/openai",
      "openai_compatible": true
    },
    "openai": {
      "env": [
        "OPENAI_API_KEY"
      ],
      "base_url": "https://api.openai.com/v1",
      "openai_compatible": true
    },
    "xai": {
      "env": [
        "XAI_API_KEY"
      ],
      "base_url": "https://api.x.ai/v1",
      "openai_compatible": true
    },
    "openrouter": {
      "env": [
        "OPENROUTER_API_KEY"
      ],
      "base_url": "https://openrouter.ai/api/v1",
      "openai_compatible": true,
      "extra_headers": {
        "HTTP-Referer": "https://claude-code.local",
        "X-Title": "claude-code-llm-harness"
      }
    }
  },
  "models": {
    "deepseek-v4-flash": {
      "provider": "deepseek",
      "cost_in": 0.14,
      "cost_out": 0.28,
      "context": 1000000,
      "speed": "fast",
      "trust": "execution",
      "good_for": [
        "code",
        "verify",
        "extract",
        "draft",
        "summarize",
        "cheap"
      ]
    },
    "deepseek-v4-pro": {
      "provider": "deepseek",
      "cost_in": 0.435,
      "cost_out": 0.87,
      "context": 1000000,
      "speed": "slow",
      "trust": "execution",
      "good_for": [
        "reason",
        "verify",
        "analyze",
        "code"
      ]
    },
    "kimi-k2.6": {
      "provider": "kimi",
      "cost_in": 0.95,
      "cost_out": 4.0,
      "context": 262144,
      "speed": "medium",
      "trust": "execution",
      "good_for": [
        "code",
        "long-context",
        "reason",
        "verify"
      ],
      "force_temperature": 1
    },
    "glm-4.5-air": {
      "provider": "zai",
      "cost_in": 0.2,
      "cost_out": 1.1,
      "context": 131072,
      "speed": "fast",
      "trust": "execution",
      "good_for": [
        "summarize",
        "extract",
        "draft",
        "cheap"
      ]
    },
    "glm-4.7": {
      "provider": "zai",
      "cost_in": 0.6,
      "cost_out": 2.2,
      "context": 203000,
      "speed": "medium",
      "trust": "execution",
      "good_for": [
        "code",
        "verify",
        "reason"
      ]
    },
    "glm-5": {
      "provider": "zai",
      "cost_in": 1.0,
      "cost_out": 3.2,
      "context": 203000,
      "speed": "medium",
      "trust": "execution",
      "good_for": [
        "code",
        "reason",
        "verify",
        "analyze"
      ]
    },
    "gemini-2.5-flash-lite": {
      "provider": "gemini",
      "cost_in": 0.1,
      "cost_out": 0.4,
      "context": 1048576,
      "speed": "fast",
      "trust": "execution",
      "good_for": [
        "cheap",
        "extract",
        "draft",
        "summarize",
        "vision"
      ]
    },
    "gemini-2.5-flash": {
      "provider": "gemini",
      "cost_in": 0.3,
      "cost_out": 2.5,
      "context": 1048576,
      "speed": "fast",
      "trust": "execution",
      "good_for": [
        "summarize",
        "extract",
        "long-context",
        "verify",
        "vision"
      ]
    },
    "gemini-3-flash-preview": {
      "provider": "gemini",
      "cost_in": 0.5,
      "cost_out": 3.0,
      "context": 1048576,
      "speed": "fast",
      "trust": "execution",
      "good_for": [
        "summarize",
        "extract",
        "long-context",
        "vision",
        "verify"
      ]
    },
    "gpt-5-nano": {
      "provider": "openai",
      "cost_in": 0.05,
      "cost_out": 0.4,
      "context": 400000,
      "speed": "slow",
      "trust": "execution",
      "good_for": [
        "cheap",
        "summarize",
        "extract",
        "draft",
        "verify"
      ],
      "token_param": "max_completion_tokens",
      "force_temperature": 1
    },
    "gpt-5-mini": {
      "provider": "openai",
      "cost_in": 0.25,
      "cost_out": 2.0,
      "context": 400000,
      "speed": "fast",
      "trust": "execution",
      "good_for": [
        "code",
        "verify",
        "reason",
        "extract"
      ],
      "token_param": "max_completion_tokens",
      "force_temperature": 1
    },
    "gpt-4.1-mini": {
      "provider": "openai",
      "cost_in": 0.4,
      "cost_out": 1.6,
      "context": 1047576,
      "speed": "fast",
      "trust": "execution",
      "good_for": [
        "code",
        "extract",
        "long-context"
      ]
    },
    "grok-4.3": {
      "provider": "xai",
      "cost_in": 1.25,
      "cost_out": 2.5,
      "context": 1000000,
      "speed": "medium",
      "trust": "execution",
      "good_for": [
        "reason",
        "verify",
        "analyze"
      ]
    }
  },
  "aliases": {
    "deepseek": "deepseek-v4-flash",
    "deepseek-pro": "deepseek-v4-pro",
    "reasoner": "deepseek-v4-pro",
    "kimi": "kimi-k2.6",
    "glm": "glm-4.7",
    "glm5": "glm-5",
    "gemini": "gemini-2.5-flash",
    "flash": "gemini-3-flash-preview",
    "lite": "gemini-2.5-flash-lite",
    "cheap": "deepseek-v4-flash",
    "nano": "gpt-5-nano",
    "mini": "gpt-5-mini",
    "grok": "grok-4.3"
  },
  "tasks": {
    "code": [
      "deepseek-v4-flash",
      "deepseek-v4-pro",
      "gpt-5-mini",
      "glm-4.7",
      "kimi-k2.6"
    ],
    "code-review": [
      "deepseek-v4-flash",
      "gpt-5-mini",
      "glm-4.7",
      "glm-5",
      "kimi-k2.6"
    ],
    "verify": [
      "deepseek-v4-flash",
      "gpt-5-nano",
      "gemini-2.5-flash",
      "deepseek-v4-pro"
    ],
    "summarize": [
      "gemini-2.5-flash-lite",
      "glm-4.5-air",
      "deepseek-v4-flash",
      "gpt-5-nano"
    ],
    "extract": [
      "gemini-2.5-flash-lite",
      "deepseek-v4-flash",
      "glm-4.5-air",
      "gpt-5-nano"
    ],
    "draft": [
      "deepseek-v4-flash",
      "glm-4.5-air",
      "gpt-5-nano"
    ],
    "reason": [
      "deepseek-v4-pro",
      "gpt-5-mini",
      "grok-4.3",
      "glm-5"
    ],
    "long-context": [
      "gemini-2.5-flash",
      "deepseek-v4-flash",
      "kimi-k2.6"
    ],
    "cheap": [
      "deepseek-v4-flash",
      "gpt-5-nano",
      "gemini-2.5-flash-lite",
      "glm-4.5-air"
    ],
    "vision": [
      "gemini-2.5-flash-lite",
      "gemini-2.5-flash",
      "gemini-3-flash-preview"
    ]
  }
}