Models

Models

What a caller can actually use right now: the engine's *loaded* models (OpenAI list shape, aliases included), not its download catalog; an id from this list always serves. Public, so the catalog is knowable before signing up; it exposes model ids and their aliases and nothing else. `?all=true` returns the engine's full catalog verbatim for browsing what could be deployed, and does need a key: that body carries per-model detail including custom cards a customer added.

The only live answer to what can be called right now: an id it returns always serves. Add ?all=true for the full catalog a node could load. Approved marketplace models are listed too.

{
  "object": "list",
  "data": [
    { "id": "mlx-community/Llama-3.2-3B-Instruct-4bit", "object": "model", "aliases": ["llama-3.2-3b"] },
    { "id": "llama-3.2-3b@8bit", "object": "model", "variant_of": "llama-3.2-3b", "quantization": "8bit", "default": false,
      "resolves_to": "mlx-community/Llama-3.2-3B-Instruct-8bit" },
    { "id": "fast", "object": "model", "resolves_to": ["mlx-community/Llama-3.2-3B-Instruct-4bit", "..."],
      "serving": "mlx-community/Llama-3.2-3B-Instruct-4bit" }
  ],
  "pricing_policy": { "long_context": { "threshold_tokens": 32768, "multiplier_bps": 15000,
                      "applies_to": "prompt_tokens + max_output_tokens" }, "...": "..." }
}

Responses: 200 Successful Response, 422 Validation Error