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

# Models & Reasoning

> Choose models and reasoning effort in Parsaa

Parsaa supports models from multiple providers. The available list is served live and kept up to date, so new models appear without an app update. Switch models any time in **Settings → Models → Chat Models**, or from the model picker in the composer.

## Providers

| Provider   | Notes                                                                                                                 |
| ---------- | --------------------------------------------------------------------------------------------------------------------- |
| Anthropic  | Claude family models                                                                                                  |
| OpenAI     | GPT and reasoning (o-series) models                                                                                   |
| Google     | Gemini family models                                                                                                  |
| OpenRouter | A single key for 100+ models across providers — see [OpenRouter](/providers/openrouter)                               |
| Local      | On-device models via LM Studio, Ollama, or any OpenAI-compatible server — see [Local Models](/providers/local-models) |

<Info>
  Because the catalog is dynamic, the exact model names you see in **Chat Models** reflect what's currently offered for your plan. Use the picker in the app as the source of truth.
</Info>

## Reasoning Effort

For models that support it, you can set how much the model "thinks" before responding. Higher effort trades speed for depth.

| Effort     | Description                          |
| ---------- | ------------------------------------ |
| Minimal    | Lightest reasoning, fastest response |
| Low        | Light reasoning                      |
| Medium     | Balanced default                     |
| High       | More deliberate reasoning            |
| Extra High | Deepest reasoning                    |

Available levels depend on your plan:

| Plan         | Reasoning levels                |
| ------------ | ------------------------------- |
| Free         | Low, Medium                     |
| Pro          | Minimal, Low, Medium, High      |
| Lite / Elite | All five (Minimal → Extra High) |

<Tip>
  Set the effort from the composer before sending. Toggle the visible thinking display from the chat footer to watch the model reason in real time.
</Tip>

## Choosing a Model

<AccordionGroup>
  <Accordion title="Quick edits and chat">
    A fast model at Low or Medium effort — ideal for simple questions, small edits, and exploration.
  </Accordion>

  <Accordion title="Complex refactoring">
    A strong coding model at Medium or High effort — better at large codebases and multi-file changes.
  </Accordion>

  <Accordion title="Deep analysis">
    A frontier model at High or Extra High effort — best for architecture decisions and intricate debugging.
  </Accordion>

  <Accordion title="Maximum privacy">
    A **Local** model — all inference runs on your Mac with no network calls. See [Local Models](/providers/local-models).
  </Accordion>
</AccordionGroup>