Choose AI Models
The AI Behavior page lets you see every model available on the platform and set the company-wide default that new employees start with, along with other workspace-wide behavior settings.
TL;DR
Go to Settings > Technical > AI Behavior to browse all available models organized by tier (Fast, Standard, Advanced). Set the company-wide default model here. The tier determines how many credits each message costs: Fast is cheapest, Advanced is most powerful. Individual employees can be switched to a different model on their profile.
Where to Find It
- Click the gear icon in the top-right of your workspace
- Go to Technical > AI Behavior
Model Tiers
Every model belongs to one of three tiers. The tier determines cost and capability.
| Tier | Speed | Capability | Credit Cost | Best For |
|---|---|---|---|---|
| Fast | Quickest responses | Good for simple, well-defined work | Lowest (roughly 10x cheaper than Advanced) | Quick lookups, formatting, simple drafts, routine Q&A |
| Standard | Balanced | Handles most everyday tasks well | Medium | Conversations, analysis, writing, planning |
| Advanced | Slower, more thorough | Best reasoning and nuance | Highest | Complex analysis, long documents, multi-step reasoning |
Available Models
Fast Tier
| Model | Provider | Credits per 1M tokens | Reasoning |
|---|---|---|---|
| GPT-5 Nano | OpenAI | 33 | Adjustable |
| GPT-4.1 Nano | OpenAI | 40 | Basic |
| GPT-4o Mini | OpenAI | 61 | Basic |
| GPT-4.1 Mini | OpenAI | 162 | Basic |
| GPT-5.4 Nano | OpenAI | 110 | Adjustable |
| GPT-5 Mini | OpenAI | 165 | Adjustable |
Standard Tier
| Model | Provider | Credits per 1M tokens | Reasoning |
|---|---|---|---|
| GPT-5.4 Mini | OpenAI | 399 | Adjustable |
| o4 Mini | OpenAI | 445 | Adjustable |
| Haiku 4.5 | Anthropic | 468 | Adjustable |
| GPT-4o | OpenAI | 1,011 | Basic |
Advanced Tier
| Model | Provider | Credits per 1M tokens | Reasoning | Notes |
|---|---|---|---|---|
| Kimi K2.5 | DeepInfra | 230 | Built-in | Recommended. Best value in Advanced tier |
| Kimi K2.6 | DeepInfra | 421 | Built-in | Newer K2 generation |
| GPT-5.1 | OpenAI | 824 | Adjustable | |
| GPT-5.2 | OpenAI | 1,154 | Adjustable | |
| Sonnet 4.6 | Anthropic | 1,404 | Adjustable | |
| GPT-5.4 | OpenAI | 2,181 | Adjustable | |
| Opus 4.7 | Anthropic | 2,340 | Adjustable | Most capable Anthropic model |
| Opus 4.6 | Anthropic | 2,340 | Adjustable | Prior Opus generation |
| GPT-5.2 Pro | OpenAI | 13,851 | Adjustable | Premium |
| GPT-5.4 Pro | OpenAI | 26,170 | Adjustable | Premium |
What You See on the Page
The AI Behavior page displays all available models grouped by tier in the Employee Brain card. For each model you can see:
- Model name and provider
- Tier (Fast, Standard, or Advanced)
- Reasoning support label, if the model supports extended thinking
- Default badge, if it is the current company-wide default
Setting the Company-Wide Default
The company-wide default model is what every newly hired employee uses unless you change it individually.
- Go to Settings > Technical > AI Behavior
- In the Employee Brain card, pick the model you want as the default
- The change saves automatically
Important: Changing the default only affects employees hired after the change. Existing employees keep their current model. To change an existing employee's model, go to their profile.
Reasoning Models
Some models support reasoning, which means the employee can think through complex problems step by step before responding. Reasoning models:
- Show a "thinking" phase in the chat before the final answer
- Produce more thorough, accurate responses for complex tasks
- Use more credits per message because of the additional thinking
Models with reasoning support can be tuned via the Thinking Depth selector in the Employee Brain card on the AI Behavior page. You can see the reasoning process in the Activity Inspector when inspecting a message.
How Models Affect Credits
The model tier is the biggest factor in credit cost per message.
| What Changes | Impact |
|---|---|
| Fast tier model | Fewest credits per message. Great for high-volume, simple work |
| Standard tier model | Moderate credits. The default balance of cost and capability |
| Advanced tier model | Most credits per message. Reserve for work that truly needs it |
| Reasoning enabled | Additional credits for the thinking phase |
You can mix and match across your workforce: assign Fast models to employees doing routine work and Advanced models to employees handling complex analysis.
Per-Employee Model Override
The company default is just the starting point. You can change any individual employee's model.
- Go to your Workspace and select the employee
- Open the Settings tab (or persona settings)
- Change the AI model
This override only affects that one employee. All other employees continue using their assigned models.
Tips and Tricks
- Start with Standard as the default. It handles most tasks well at a reasonable cost. Switch individual employees to Fast or Advanced based on their specific work
- Use Fast for high-volume employees. If an employee answers lots of simple questions or does routine formatting, Fast saves significant credits
- Reserve Advanced for where it matters. Complex reasoning, long document analysis, and nuanced decision-making benefit most from Advanced models
- Check reasoning support. If an employee needs to work through multi-step problems, assign a model with reasoning support
Frequently Asked Questions
Q: Where do I set the company-wide default model? A: Go to Settings (gear icon) > Technical > AI Behavior. In the Employee Brain card, pick the model you want. It saves automatically.
Q: Does changing the default model affect existing employees? A: No. Only newly hired employees use the new default. Change existing employees individually on their profile.
Q: What is the difference between tiers? A: Fast is cheapest and quickest for simple tasks. Standard balances cost and capability for everyday work. Advanced is most powerful for complex reasoning but costs the most credits.
Q: Can different employees use different models? A: Yes. The company default is just the starting point. You can assign any available model to any individual employee on their profile.
Q: Do reasoning models cost more? A: Yes. The thinking phase uses additional credits. The total cost depends on how much reasoning the model does for each response.