> ## Documentation Index
> Fetch the complete documentation index at: https://docs.apiyi.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Why Does Hermes Show Only Some Models in Its Model List?

> Learn why the Hermes model picker has a 50-model limit per provider and how to switch to models outside the visible list by ID.

## Short answer

In the Hermes implementation reviewed for this case, the model picker shows at most **50 models per provider**. This is a limit on the data sent to the picker, not a limit on the API key's model permissions. An API endpoint or Hermes cache can contain more model IDs; models outside the visible list can still be selected by their full ID, though each model must be tested separately.

The picker truncates the `models` field to `model_ids[:max_models]`, while `total_models` retains the count before truncation. So `50 available` means the menu displays at most 50 entries; it does not mean the provider has only 50 models.

Hermes Desktop and the terminal picker should use the same selection logic when they share a backend, profile, provider, and model catalog. If their counts differ, the 50-item cap explains why a picker shows fewer models than the full catalog, but first make sure both views are for the same provider and catalog.

## Understand the three lists

| Entry point | What it represents |
| - | - |
| `/v1/models` or `provider_models_cache.json` | Model IDs returned by the API or stored in Hermes' cache |
| Hermes Desktop / `/model` picker | At most 50 model IDs per provider |
| `/model <model-id>` | Switches the current chat to a full model ID without requiring it to appear in the first 50 |

## Troubleshooting steps

1. **Check the provider row**: `openai-api`, `openai`, and `openrouter` are separate provider entries, and the limit applies to each provider independently. Confirm that Desktop and CLI are showing the same row instead of comparing one provider's selectable count with another provider's catalog total.
2. **Enter the model ID directly if it is missing**: In a chat, use `/model gpt-5.5`, or start a session with `hermes chat --model gpt-5.5`. Replace `gpt-5.5` with the exact model ID returned by the API endpoint. Add a provider prefix if Hermes requires one for the current setup.
3. **Set and verify the default model**: Use `hermes config set model.default gemini-3.8-flash`, then run `hermes config show` to confirm that the value was written to the right field. `hermes config set` does not validate whether the model name is available; verify it with an actual request.
4. **Separate the list cap from call failures**: Seeing 50 entries only means the picker truncated the menu; it does not prove every model after item 50 is available. Send a minimal request for the full model ID and check APIYI call logs for delivery, status code, and error details.

There is currently no `config.yaml` setting for changing the 50-item cap. If Desktop and CLI still return different counts under the same Hermes version, profile, provider, and cache, record the provider names, full model IDs, Hermes backend version, and call logs. Remove API keys before sharing logs.

## Related documentation

* [Connect Hermes Agent to APIYI](/en/scenarios/agent/hermes-agent)
* [How do I choose the right AI model?](/en/faq/model-selection-guide)
* [Why can't I use some models?](/en/faq/model-availability)
* [How do I view my call records?](/en/faq/call-logs)

For Hermes' official references, see `hermes-agent.nousresearch.com/docs/user-guide/desktop`, `hermes-agent.nousresearch.com/docs/user-guide/configuring-models`, and `hermes-agent.nousresearch.com/docs/reference/model-catalog`.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.