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

# Choose AI Models

> Learn how to choose the right AI models for your project

Pick a model that matches your task. You can switch models at any time before sending a message.

<Frame caption="Select a model before you send your message">
  <img src="https://mintcdn.com/steercode/l9csF7jAQiR_B1h5/images/select-model-web.png?fit=max&auto=format&n=l9csF7jAQiR_B1h5&q=85&s=66026127d8be9733bcf8fda666906f91" alt="Model picker in the SteerCode composer" width="1050" height="794" data-path="images/select-model-web.png" />
</Frame>

### Available models

| Model                     | Provider  | Profile                      | Thinking/Reasoning          | Credit usage           |
| ------------------------- | --------- | ---------------------------- | --------------------------- | ---------------------- |
| **GPT-5**                 | OpenAI    | Most advanced coding model   | ✅ Deep multi-step reasoning | **Premium (credits)**  |
| **GPT-4.1**               | OpenAI    | Fast senior programmer       | –                           | **Premium (credits)**  |
| **GPT-4.1 mini**          | OpenAI    | Fast junior programmer       | –                           | **Basic (no credits)** |
| **Gemini 2.0 Flash**      | Google    | Fast junior programmer       | –                           | **Basic (no credits)** |
| **Claude Sonnet 3.7 / 4** | Anthropic | Strong quality-speed balance | ✅                           | **Premium (credits)**  |
| **Claude Opus 4**         | Anthropic | Highest-quality reasoning    | ✅                           | **Premium (credits)**  |

<Info>
  You may occasionally see **experimental / stealth models** (e.g., *Horizon
  Alpha*). Availability changes—watch our socials for announcements.
</Info>

<Tip>
  If you’re iterating rapidly or doing light edits, choose a **Basic** model to
  **avoid credit usage**. Switch to a **Premium** model for complex refactors,
  tricky bugs, or multi-step integrations.
</Tip>

### Quick guidance

* **Speed first** (UI copy, small edits, quick prompts): *GPT-4.1 mini* or *Gemini Flash 2.0*
* **General coding** (features, refactors, tests): *GPT-4.1* or *Claude Sonnet*
* **Hard tasks** (deep reasoning, multi-file migrations): *GPT-5* or *Claude Opus*

<Warning>
  Thinking-enabled models (e.g., GPT-5, Sonnet/Opus) are slower and consume more
  credits. Use them when the extra reasoning matters.
</Warning>