In the Outmind Assistant, you can choose the AI model that handles your request and its reasoning level: a quick answer for everyday requests, or a more in-depth analysis for complex tasks.
⚠️ Prerequisite: the models offered depend on the configuration chosen by your Outmind administrator. If a model doesn't appear in the list, contact your administrator (see the article Choosing your teams' AI models (Admin)).
🤖 1. Choose your model
In the Assistant:
Below the message box, click the button showing the current model (for example "Claude Opus 5.5 Low").
Click the Model line.
Select the model you want.
Choose your model according to your needs : the more powerful it is, the better it handles complex tasks.
Model | Provider | Cost | Best for | Reasoning levels |
Claude Haiku 5.5 | Anthropic | € | The fastest and most cost-effective, for everyday tasks. | ✅ Yes |
Claude Sonnet 5.5 | Anthropic | €€ | The balance between performance, speed and cost. | ✅ Yes |
Claude Opus 5.5 | Anthropic | €€€ | Complex reasoning and long analyses across several documents. | ✅ Yes |
🧠 2. Adjust the reasoning level
In the Assistant:
Below the message box, click the button showing the current model and reasoning level.
Drag the Reasoning slider towards "Faster" for an immediate answer, or towards "Smarter" for an in-depth analysis.
Higher reasoning lets you handle more complex tasks.
Your task | Examples | Recommended reasoning |
✉️ Proofreading / rewording | Emails and short texts | 🟢 Low |
📝 Writing / summarizing | Meeting notes and summaries | 🟢 Low to 🟡 medium |
📄 Document analysis | Extracting and comparing documents | 🟡 Medium to 🔴 high |
🔍 Complex analysis | DCE, risks and inconsistencies | 🔴 High |
⚠️ A higher reasoning level can improve a difficult analysis, but it can also make responses slower. You don’t need the maximum level for every request: the right choice mainly depends on the task. There’s no need for the maximum setting to proofread a simple email.
💡 Good habit: start with low reasoning, and increase it only if the answer isn't enough. The higher the reasoning, the longer the answer takes.


