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

# AI Models

> UltraClaw supports Claude, GPT, and Gemini models. Switch models anytime from your dashboard or in chat.

# AI Models

Your UltraClaw agent supports multiple AI models from the leading providers. Switch between them anytime to find the right balance of speed, capability, and cost for the task at hand.

## Supported Models

<CardGroup cols={3}>
  <Card title="Claude">
    <img src="https://mintcdn.com/ultraclaw/vfYrk54kefzlWrdf/images/claude.svg?fit=max&auto=format&n=vfYrk54kefzlWrdf&q=85&s=e04755bba70d62374ac8177e9f4a17cf" alt="Anthropic Claude" style={{ height: "32px", marginBottom: "8px" }} width="1200" height="1200" data-path="images/claude.svg" />

    **Anthropic** — excellent at writing, analysis, nuanced reasoning, and following complex instructions.
  </Card>

  <Card title="GPT">
    <img src="https://mintcdn.com/ultraclaw/vfYrk54kefzlWrdf/images/chatgpt.png?fit=max&auto=format&n=vfYrk54kefzlWrdf&q=85&s=c1d221bd681ace4039b79f22f083e892" alt="OpenAI GPT" style={{ height: "32px", marginBottom: "8px" }} width="512" height="512" data-path="images/chatgpt.png" />

    **OpenAI** — strong general-purpose capabilities with fast response times and broad knowledge.
  </Card>

  <Card title="Gemini">
    <img src="https://mintcdn.com/ultraclaw/vfYrk54kefzlWrdf/images/gemini.png?fit=max&auto=format&n=vfYrk54kefzlWrdf&q=85&s=bdae25f03adfb462e24b5987411ea584" alt="Google Gemini" style={{ height: "32px", marginBottom: "8px" }} width="960" height="960" data-path="images/gemini.png" />

    **Google** — powerful multimodal understanding with access to Google's latest research.
  </Card>
</CardGroup>

Your agent always uses the latest available versions of each model family. As providers release new models, your agent is updated automatically.

## Switching Models

You can change your agent's model in two ways:

<Tabs>
  <Tab title="In Chat">
    Use the `/new` slash command followed by the model name to start a new session with a different model:

    ```
    /new claude
    /new gpt
    /new gemini
    ```

    This starts a fresh session with the selected model. Your memory and workspace files carry over — only the model changes.
  </Tab>

  <Tab title="Dashboard">
    Open your [dashboard](https://ultraclaw.ai/dashboard), go to **Settings**, and select your preferred default model. This model will be used for all new sessions.
  </Tab>
</Tabs>

<Tip>
  You can switch models mid-workflow. Stuck on a problem with one model? Try `/new claude` or `/new gpt` for a fresh perspective.
</Tip>

## Thinking Levels

Control how deeply your agent reasons through problems with the `/think` command. Higher thinking levels produce more thorough analysis but consume more tokens.

| Level          | Command           | Best For                                           |
| -------------- | ----------------- | -------------------------------------------------- |
| **Off**        | `/think off`      | Quick lookups, simple questions                    |
| **Minimal**    | `/think minimal`  | Straightforward tasks, short answers               |
| **Low**        | `/think low`      | Everyday conversations and requests                |
| **Medium**     | `/think medium`   | Moderate reasoning, planning                       |
| **High**       | `/think high`     | Complex analysis, multi-step problems              |
| **Extra High** | `/think xhigh`    | Deep research, intricate reasoning                 |
| **Adaptive**   | `/think adaptive` | Let the agent choose the right level automatically |

<Info>
  **Adaptive** is the default thinking level. Your agent automatically adjusts its reasoning depth based on the complexity of your request — using more tokens only when the task demands it.
</Info>

## Model Fallback

If your selected model is temporarily unavailable due to provider rate limits, your agent **automatically rotates** to another available model. You'll receive a seamless response without needing to do anything.

Once the preferred model is available again, your agent switches back automatically.

## Token Costs by Model

Different models consume tokens at different rates. In general:

* **Lighter models** are more cost-efficient for simple tasks
* **Premium models** cost more per token but deliver stronger reasoning

All plans include **\$15/mo in AI credits**, with top-up available anytime. See [Tokens & Usage](/plans/tokens-and-usage) for details on how credits work.

## Choosing the Right Model

| Task                         | Recommended Approach                               |
| ---------------------------- | -------------------------------------------------- |
| Quick questions & lookups    | Any model with `/think off` or `/think minimal`    |
| Writing & editing            | Claude tends to excel at nuanced writing           |
| Data analysis & coding       | GPT and Claude both perform well                   |
| Complex multi-step reasoning | Any model with `/think high` or `/think xhigh`     |
| General everyday use         | Use **Adaptive** thinking and your preferred model |