Models

GET /v1/models

Lists available models in the OpenAI shape:

curl "$SONIX_URL/v1/models" -H "Authorization: Bearer $SONIX_API_KEY"

{
  "object": "list",
  "data": [
    {
      "id": "kokoro-82m",
      "object": "model",
      "created": 1710000000,
      "owned_by": "sonix"
    }
  ]
}

Capability routing

Every model has fixed capabilities. Endpoints only accept models that match. A speech model works on /v1/audio/speech; a transcription model works on /v1/audio/transcriptions; and so on. The ElevenLabs routes reach only the deployment's configured speech model — see below.

Sending a model to an incompatible endpoint is rejected with a validation error naming param: "model" once the model's capabilities are known to the serving instance. Capabilities are read from the model bundle, so an instance that has not yet loaded that model cannot check the match up front and the request fails later instead. See Errors.

The ElevenLabs routes don't take a model parameter; they route to the deployment's configured speech model.

Beyond voice, the same API serves chat completions (/v1/chat/completions) when matching models are deployed.