List models

Return every upstream model routeur.ai can currently route to. OpenAI-compatible: callers built against GET /v1/models work without changes, which is what lets tools that probe this endpoint before they will talk to a base URL at all — the OpenAI SDKs, LangChain, LibreChat, Open WebUI, Continue and the editor integrations — point at routeur.ai unmodified.

GET /v1/models Stable
https://api.routeur.ai

Query parameters

provider string optional
Restrict the result to one provider, e.g. openai.
configured boolean optional
Restrict to models whose provider credentials are currently mounted.

Returns

An object with object: "list" and data (array of Model objects).

Each entry carries OpenAI's fields (id, object, created, owned_by) so generic clients parse it, plus routeur.ai's own (provider, context_window, pricing, configured, aliases).

created is present for schema compatibility but is a fixed constant, not a release date — routeur.ai's catalogue tracks pricing and context windows, not model provenance. It is stable between calls, so a client that sorts or diffs on it behaves predictably; it just does not tell you when a model shipped.

id is unique within a response. Where two providers serve the same model id, the listing shows the one a pinned call to that id would actually route to.

What you see

The listing is your workspace's, not a global vendor catalogue: it contains the models that workspace has activated, so two organisations calling this endpoint legitimately get different answers. auto is listed first — it is a real value you can send in the model field, and it asks the gateway to choose.

configured reports whether a credential for that model's provider is currently mounted. An activated model whose provider key has been removed is listed with configured: false rather than hidden, so a failed call has a visible explanation. The ?configured= query parameter filters on it in both directions — true for routable models, false for exactly the ones that need a credential adding.

This listing covers chat models. Embedding models are resolved from a separate catalogue and are not activated per workspace, so they do not appear here — see Create embeddings for the models that endpoint accepts.

This endpoint is not gated by billing or usage state — a capped or lapsed organisation can still list its models, because a client that cannot connect at all cannot discover why it is being refused.