Models and identifiers

Updated

Requests use three kinds of identifier: the model ID decides which model runs, and the provider and endpoint identifiers decide who serves it.

Model ID

The model ID of each model in the catalog is the value of the model field, usually publisher/model-name, for example deepseek/deepseek-v3.2.

  • A model alias works too; the model in the response is always the canonical model ID.
  • The Gemini API takes the model ID in the path: /v1beta/models/deepseek/deepseek-v3.2:generateContent, and the slash may be written as %2F.
  • A model in maintenance or that does not exist returns 404 model_not_found; a retired model returns 410 model_retired. See Model capabilities for statuses.

Provider and endpoint identifiers

One model can have offers from several providers. Each endpoint is one provider serving the model at one price with one configuration.

Identifier Example Use
Provider identifier Shown in the endpoint list on the model page Name all of a provider's endpoints in order, only or ignore of provider
Endpoint identifier Starts with route_ Lock a request to exactly one endpoint with only; see Provider selection

Querying from code

Endpoint Auth Returns
GET https://modelportal.ai/v1/models Key required OpenAI format; only models this key may call that have an available endpoint
GET https://modelportal.ai/api/v1/models None Platform-format model list: capabilities, status, price summary, provider count, and each endpoint's configuration, price and rate limits
GET https://modelportal.ai/api/v1/models/<model ID> None One model in detail, with each endpoint's identifier, provider, price, data policy and supported parameters; write the / in the model ID as %2F

See the Models API for the fields.