Reasoning Effort
Reasoning effort sets how much the model thinks before it answers. Higher effort gives more thorough answers on complex tasks but takes longer and uses more tokens. Lower effort is faster and cheaper.
You pick the effort in the chat itself, so each chat can run at its own level.
Prerequisites
An LLM provider configured by an account Owner.
Change the Effort
The effort control sits at the bottom right of the message box in every agent chat: Onboarding, Schema & Relationships, Business Requirements, AI Modeling, dbt and Ask.
- Click the current level (for example Extra high) next to the send button
- The menu shows Effort · model name with the levels that model supports
- Click a level: Low, Med, High, XHigh or Max
The dot marks the model's default level. The new level applies from your next message.
How It Works
| Behavior | Details |
|---|---|
| Per chat | The level is saved for the chat you set it in. Other chats keep their own level. |
| Main model only | Effort applies to the model that answers in the chat. Background tasks, such as summarizing long conversations, keep their default settings. |
| During a response | The control is disabled while the agent is responding. |
| Inactive chats | A chat with no activity for a day returns to the default level. |
| Unsupported level | If the Owner switches to a model that does not support the saved level, the chat runs at that model's default. |
| No effort setting | Models without an effort setting do not show the control. |
Levels by Model
| Models | Levels | Default |
|---|---|---|
| Claude Opus 5.5, Claude Sonnet 5.5 (Anthropic, Vertex AI) | Low, Med, High, XHigh, Max | XHigh |
| Claude Haiku 5.5 (Anthropic, Vertex AI) | Low, Med, High, XHigh, Max | High |
| GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Luna (OpenAI, Azure OpenAI) | Low, Med, High, XHigh, Max | High |
| Gemini 4 Argon, Gemini 3.8 Flash (Gemini, Vertex AI) | Low, Med, High | High |
| Claude Opus 5, Claude Sonnet 5 (AWS Bedrock) | Low, Med, High, XHigh, Max | XHigh |
| Claude Opus 5, Claude Sonnet 5 (Snowflake Cortex) | Low, Med, High, Max | High |
| Claude Haiku 4.5 (AWS Bedrock, Snowflake Cortex) | No effort setting | — |
Next Steps
- LLM Configuration - Choose the provider and models
- Onboarding Agent - Start a chat on your data