Choosing a model
What Auto does, how the models differ, and what each one costs.
Open the model menu next to the prompt box. Every model shows its quality, speed and cost.
Auto
Auto uses the lowest-cost model (DeepSeek V4.1 Flash first, then GPT-6 Luna, Claude Haiku 5.5, DeepSeek V4 Pro and Gemini 3.8 Flash). If that model is busy or fails, Auto tries the next one, and you're only charged for the one that answers. Pick a model yourself when you want a stronger one.
Picking a model yourself
When you pick a model, that model always answers. It's never swapped for another. If it's busy or unavailable you'll see a message, and nothing is charged.
The lineup
| Model | Maker | Quality | Speed | Credits (minimum) |
|---|---|---|---|---|
| Claude Fable 5.1 | Anthropic | Best | Thorough | 10 |
| Claude Opus 5.5 | Anthropic | Best | Thorough | 5 |
| Claude Sonnet 5.5 | Anthropic | Great | Fast | 2 |
| Claude Haiku 5.5 | Anthropic | Good | Fastest | 1 |
| GPT-6 Astra | OpenAI | Best | Thorough | 10 |
| GPT-6.1 Sol | OpenAI | Great | Fast | 2 |
| GPT-6 Luna | OpenAI | Good | Fastest | 1 |
| Gemini 4 Argon | Great | Fast | 2 | |
| Gemini 3.8 Flash | Good | Fastest | 1 | |
| Muse Wonder 1.0 | Meta | Best | Thorough | 5 |
| Kimi K3 | Moonshot | Best | Thorough | 4 |
| Qwen 3.8 Max Prime | Alibaba | Great | Fast | 3 |
| DeepSeek V4 Pro | DeepSeek | Great | Fast | 1 |
| DeepSeek V4.1 Flash | DeepSeek | Great | Fastest | 1 |
| GLM 5.3 | Z.ai | Great | Fast | 1 |
| Mistral Large 4 | Mistral | Great | Fast | 1 |
Which should I use?
- Everyday building: Claude Sonnet 5.5, GPT-6.1 Sol or Gemini 4 Argon.
- Quick, low-cost tweaks: Claude Haiku 5.5, GPT-6 Luna, Gemini 3.8 Flash or DeepSeek V4.1 Flash (a fraction of a credit for most changes).
- Big, complex projects: Claude Fable 5.1 or GPT-6 Astra, which cost more credits. Use them for the hard parts.
Every model needs a Pro or Max plan. See Plans and payments.
Related articles
Didn't find what you needed? Contact support