Advanced AI model settings
Advanced AI model settings
Temperature
Temperature controls response creativity:
- 0 to 0.3: factual and deterministic responses (recommended for analysis)
- 0.5 to 0.7: creativity/precision balance (general use)
- 0.8 to 1.0: creative and varied responses (brainstorming)
Max tokens
Sets the maximum response length. A higher value allows longer responses but consumes more credits.
Top P (Nucleus sampling)
Controls vocabulary diversity. Recommended value: 0.9.
System prompt
In the Agents app, you can define a system prompt to customize model behavior.
Model selection
| Model | Recommended use |
|---|---|
| GPT-4o | General use, complex analysis |
| Claude 3.5 | Long-form writing, document analysis |
| Mistral Large | Sovereignty, European use cases |
| Gemini | Multimodal, web search |