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

# Configure Models

> Choose the models and reasoning levels a router may select.

The default `/v1` router starts with these Dari-managed models:

```text theme={null}
anthropic/claude-fable-5
openai/gpt-5.6-sol
fireworks/zai-org/GLM-5.2
fireworks/deepseek-ai/DeepSeek-V4-Flash
```

Edit the default router in the dashboard to change its models, or create another router and set it as the default.

## Public Models

Add public models from the router's **Models & Keys** step. Model IDs include their provider:

```text theme={null}
openai/gpt-5.6-sol
anthropic/claude-sonnet-4-6
fireworks/deepseek-ai/DeepSeek-V4-Pro
```

The provider prefix determines which [provider key](/router/configure-provider-keys) Dari uses. Run `dari router models` or use the dashboard picker to see the current catalog.

## Custom Models

To route to an OpenAI-compatible Chat Completions model that is not in the public catalog, open **Models** in the dashboard and select **Add Model**. Register its HTTPS base URL, the exact model ID its endpoint expects, maximum input-plus-output token count, prices in dollars per million tokens, and an organization credential. The model then appears under **Your Models** when you create or edit a router.

Custom models are private to your organization. Their saved credential is used for upstream requests, so they do not need a separate router-level provider key.

## Reasoning Levels

By default, the router may use every reasoning level supported by each model. Restrict them in YAML when needed:

```yaml theme={null}
enabled_models:
  - openai/gpt-5.6-sol
  - openai/gpt-4.1-mini
model_thinking_levels:
  openai/gpt-5.6-sol: [low, medium, high]
  openai/gpt-4.1-mini: [off]
```

Supported level names are `off`, `minimal`, `low`, `medium`, `high`, `xhigh`, and `max`.