models_catalog.json

{
  "//": "Logical-model catalog: a stable logical name -> ORDERED (provider, model)",
  "//1": "candidates. route_judge tries candidates best-first (live-verified slots are",
  "//2": "ordered first) and bounces to the next on any failure. Curated 2026-06 from",
  "//3": "live testing. Tiers below are organisational only; any request can name any",
  "//4": "logical model or a physical 'provider/model' slot directly.",

  "logical_models": {
    "kimi-k2.6": {
      "family": "kimi-k2",
      "candidates": [
        {"provider": "nvidia", "model": "moonshotai/kimi-k2.6", "ctx": 262144},
        {"provider": "cloudflare", "model": "@cf/moonshotai/kimi-k2.6", "ctx": 262144},
        {"provider": "openrouter", "model": "moonshotai/kimi-k2.6:free", "ctx": 262144}
      ]
    },
    "deepseek-v4-flash": {
      "//": "Primary DeepSeek (2026-06): v4-pro had an unusable >1k queue (300s ReadTimeout).",
      "//1": "v4-flash is the 284B MoE distillation, ~5-8% below pro but reliably served.",
      "family": "deepseek-v4-flash",
      "candidates": [
        {"provider": "nvidia", "model": "deepseek-ai/deepseek-v4-flash", "ctx": 262144},
        {"provider": "github-models", "model": "deepseek/deepseek-v3-0324", "ctx": 8000}
      ]
    },
    "deepseek-v4-pro": {
      "//": "DEMOTED 2026-06: queue >1k, ReadTimeout-prone. Prefer deepseek-v4-flash.",
      "family": "deepseek-v4",
      "candidates": [
        {"provider": "nvidia", "model": "deepseek-ai/deepseek-v4-pro", "ctx": 163840}
      ]
    },
    "deepseek-v3": {
      "family": "deepseek-v3",
      "candidates": [
        {"provider": "github-models", "model": "deepseek/deepseek-v3-0324", "ctx": 8000}
      ]
    },
    "glm-5.1": {
      "family": "glm",
      "candidates": [
        {"provider": "nvidia", "model": "z-ai/glm-5.1", "ctx": 131072}
      ]
    },
    "nemotron-ultra": {
      "family": "nemotron-3",
      "candidates": [
        {"provider": "nvidia", "model": "nvidia/nemotron-3-ultra-550b-a55b", "ctx": 131072},
        {"provider": "openrouter", "model": "nvidia/nemotron-3-ultra-550b-a55b:free", "ctx": 131072}
      ]
    },
    "deepseek-r1": {
      "family": "deepseek-r1",
      "candidates": [
        {"provider": "github-models", "model": "deepseek/DeepSeek-R1-0528", "ctx": 8000}
      ]
    },
    "nemotron-120b": {
      "family": "nemotron-3-120b",
      "candidates": [
        {"provider": "cloudflare", "model": "@cf/nvidia/nemotron-3-120b-a12b", "ctx": 262144}
      ]
    },

    "llama-3.3-70b": {
      "family": "llama-3.3-70b",
      "candidates": [
        {"provider": "cloudflare", "model": "@cf/meta/llama-3.3-70b-instruct-fp8-fast", "ctx": 24000},
        {"provider": "github-models", "model": "meta/llama-3.3-70b-instruct", "ctx": 8000}
      ]
    },
    "gpt-oss-120b": {
      "family": "gpt-oss",
      "candidates": [
        {"provider": "cloudflare", "model": "@cf/openai/gpt-oss-120b", "ctx": 128000}
      ]
    },
    "phi-4-reasoning": {
      "family": "phi-4",
      "candidates": [
        {"provider": "github-models", "model": "microsoft/Phi-4-reasoning", "ctx": 8000}
      ]
    },
    "step-flash": {
      "family": "step-3.7",
      "candidates": [
        {"provider": "nvidia", "model": "stepfun-ai/step-3.7-flash", "ctx": 131072}
      ]
    },
    "gemma-4": {
      "family": "gemma-4",
      "candidates": [
        {"provider": "nvidia", "model": "google/gemma-4-31b-it", "ctx": 131072},
        {"provider": "openrouter", "model": "google/gemma-4-31b-it:free", "ctx": 96000},
        {"provider": "cloudflare", "model": "@cf/google/gemma-4-26b-a4b-it", "ctx": 262144}
      ]
    },

    "//roster-2026-06": "Additions below all live-probed OK (docs/ROSTER-2026-06.md); all stronger than gemma-4.",
    "qwen3.5-397b": {
      "family": "qwen3.5-397b",
      "candidates": [
        {"provider": "nvidia", "model": "qwen/qwen3.5-397b-a17b", "ctx": 262144}
      ]
    },
    "qwen3.5-122b": {
      "family": "qwen3.5-122b",
      "candidates": [
        {"provider": "nvidia", "model": "qwen/qwen3.5-122b-a10b", "ctx": 262144}
      ]
    },
    "llama-4-maverick": {
      "family": "llama-4-maverick",
      "candidates": [
        {"provider": "nvidia", "model": "meta/llama-4-maverick-17b-128e-instruct", "ctx": 262144}
      ]
    },
    "minimax-m2.7": {
      "family": "minimax-m2",
      "candidates": [
        {"provider": "nvidia", "model": "minimaxai/minimax-m2.7", "ctx": 262144}
      ]
    },
    "mistral-medium-3.5": {
      "family": "mistral-medium-3.5",
      "candidates": [
        {"provider": "nvidia", "model": "mistralai/mistral-medium-3.5-128b", "ctx": 131072}
      ]
    },
    "mistral-small-4": {
      "family": "mistral-small-4",
      "candidates": [
        {"provider": "nvidia", "model": "mistralai/mistral-small-4-119b-2603", "ctx": 131072}
      ]
    }
  }
}