Back to tips

Control AI Thinking Effort Levels

Adjust how deeply your AI agent reasons through complex problems

ai intermediate February 26, 2026 · godruoyi
#agent #ai #performance #configuration

When using the Zed provider with supported models, you can now control thinking effort levels to balance between response speed and reasoning depth.

How to Use

Thinking effort controls appear in the agent panel when using compatible models:

  1. Open an agent thread
  2. Look for the thinking effort selector (appears with supported models)
  3. Choose your preferred level based on task complexity
  4. The agent will adjust its reasoning depth accordingly

Higher effort levels provide more thorough analysis at the cost of longer response times.

Why This Matters

Task Optimization: Use lower effort for simple questions, higher effort for complex reasoning tasks.

Cost Control: Higher thinking effort may consume more tokens, so you can optimize based on your needs.

Response Speed: Get faster responses when deep reasoning isn’t necessary.

This feature is particularly useful when you need quick answers for straightforward questions but want deeper analysis for architectural decisions or complex debugging.