Skip to main content

​
What you saw

The model name varies based on the model value your client sent.

​
Why it happened

You sent a concrete model ID instead of auto to an OpenAI-format endpoint, but that model is not available to this harness. This can happen when:
  1. The model ID is misspelled or retired.
  2. The provider is not connected yet, or it is connected but not enabled for this harness.
  3. You used a provider-native name when Manifest expects the exact ID returned by GET /v1/models.
  4. Manifest refuses the request instead of falling back to automatic routing.

​
How to fix it

  1. Call GET /v1/models with the same Manifest harness key.
  2. Use auto to let Manifest route, or copy one of the listed model IDs exactly.
  3. If the model should be available, open the dashboard link in the error and enable or connect that provider for this harness.
  4. Retry the request.