Models

GET /v1/models — list models currently available in the pool.
Standard OpenAI-shape /v1/models. Returns the set of models advertised by connected providers right now — the list changes as Macs come online and offline. Authentication required. Unauthenticated calls return 401 missing_bearer_token.

Endpoint

Response

Example shape (IDs are illustrative — call the endpoint):
On 16 August 2026 those two IDs were the warm set. Do not hard-code availability.

From Python

Availability

A model appears in the list when at least one connected provider advertises it warm. If every provider serving a given model disconnects, the model drops out of the list within a heartbeat cycle. Requests for a dropped-out model return 404.

Full catalog vs available now

/v1/models shows only what’s available right now. For priced rows and the signed catalog (including models not currently loaded), see Operations → Model catalog.

Rate card

Per-model pricing:
Returns per-model prompt_rate_per_mtok, completion_rate_per_mtok, and prompt_cache_hit_rate_per_mtok, plus usd_per_million_credits. Rate-card slugs may differ from MLX serve IDs. Dated comparison: Pricing comparison.