Hermes "The model request failed. Run /model to switch or hermes doctor to check the setup."

Why it happens

The CLI chat panel maps the error classifier's verdict to plain-language copy. Anything the classifier cannot place lands on this generic message, with the raw provider text demoted to a Details line.

1. A model id the provider does not recognise or has retired, reported in words the classifier does not catch.
2. A provider-specific rejection, such as a bad tool schema or an unsupported parameter (often an HTTP 400).
3. A broken provider setup: wrong base_url, a proxy in the way, or a custom provider entry with a typo.
4. On older builds, a lost connection to the local gateway was also shown this way, which no model switch can fix.

The fix

1. 1 Read the Details line. It is the provider's real answer.
2. 2 If it names a model, pick a valid one with /model.
3. 3 If it mentions 401, 403 or an invalid key, fix the key in ~/.hermes/.env and restart Hermes.
4. 4 If it is a 400 or a schema complaint, check the model's tool and context support.
5. 5 If every model fails, run hermes doctor, and hermes gateway status to rule out a lost gateway.

Still failing?

• Run hermes logs to see the full provider response.
• Update Hermes: newer builds classify more errors and print a specific fix instead of this line.

Related errors

• Hermes "Provider authentication failed"
• Hermes "API call failed after 3 retries"