Skip to main content
Not yet available. GET /v1/models does not currently return a usable catalogue, and GET /v1/models/{model} is not implemented. Both are on the roadmap.For the list of models you can call today, see Models — it is the authoritative reference and includes context windows, capabilities, and pricing.

Current behaviour

The single "*" entry means “this gateway can route any configured model”. It is not a catalogue, and code that populates a model picker from it will find nothing usable. GET /v1/models/{model} returns 400 MISSING_CONTEXT.

What to do instead

Pin the model IDs your application uses rather than discovering them at runtime. IDs from Models are stable — a model is retired with notice, never renamed underneath you.
Send a minimal request and check for a 404:
A 404 with Model not found means the ID is not available on your account. Anything else means the model is reachable.
Some clients call /v1/models before their first request. Most treat an unrecognised response as “no models offered” and still work once you set a model explicitly in their configuration.Codex CLI is unaffected — it never calls this endpoint. See the Codex CLI guide.
When this endpoint ships it will require authentication and will list only the models your key is permitted to call — so a listing will always match what a follow-up request is allowed to do.