Models

GET /v1/models lists the models the key’s tenant can call at this moment. Tools built on the OpenAI SDK call it first to fill a model picker, so it answers in the OpenAI shape.

Request

curl -s https://<your-gateway-host>/v1/models \
  -H "Authorization: Bearer $OLYMPUS_API_KEY"

from openai import OpenAI
import os

client = OpenAI(base_url="https://<your-gateway-host>/v1", api_key=os.environ["OLYMPUS_API_KEY"])
for model in client.models.list():
    print(model.id)

Response

{
  "object": "list",
  "data": [
    {
      "id": "<model-name>",
      "object": "model",
      "created": 0,
      "owned_by": "olympus",
      "context_window": 32768,
      "supports_tools": true
    }
  ]
}

Field	Meaning
`id`	The name to send as `model`.
`object`, `owned_by`	Always `model` and `olympus`.
`context_window`	The most tokens the model takes, prompt and answer together.
`supports_tools`	Whether the model returns tool calls.

Clients that do not know the last two fields ignore them.

What is listed

A model is listed when all of these hold for the key’s tenant:

the tenant is granted it;
a deployment of it is ready, or a cloud provider serves it;
the gateway has a route for it.

A model that is granted and not yet serving is not listed, because choosing it would fail. An empty list is a valid answer: the tenant has nothing serving yet.

The list is the tenant’s, not the key’s: a key limited to some models still sees every model its tenant can call, and is refused 403 model_not_allowed for the others.