Returns the models available to the authenticated credential: the environment’s catalog, narrowed by the key’s allowed_models restriction when present. Each entry carries the model’s capabilities.

​
Authentication

Authenticated with a secret key (rd_sk_…) or a client token of any scope — the catalog is non-sensitive discovery data, already narrowed by the (inherited) key restrictions, so no per-scope gate applies. A publishable key (rd_pk_…) is mint-only and is rejected 403 (publishable_key_mint_only).

​
Response

200 OK:
  • id — the provider/model id to send as model on a request.
  • provider — the provider that serves the model (openai or anthropic).
  • display_name — a human-readable label.
  • capabilities — the model’s declared capabilities: context_window, max_output, modalities (e.g. ["text","image"]), tools, json_schema, streaming. The gateway uses these to reject requests a model cannot satisfy (e.g. tools against a model with tools: false).
This endpoint lists concrete models. Project aliases (like smart) are configured in the dashboard and are also valid model values on a request — see Models, aliases & fallback.