GitHub Copilot Cloud Agent Lets You Customize the Reasoning Level per Task
GitHub has added the ability to customize the reasoning level when delegating a task to Copilot cloud agent. Users can now set how much a supported model reasons before responding, alongside choosing the model itself.
Details
- What’s new: When starting a task with Copilot cloud agent, users can set the reasoning level for models that support it, in addition to selecting which model to use
- Trade-off: Higher reasoning levels can improve responses to complex problems, but consume more tokens and Copilot credits
- Availability: Available on all paid Copilot plans that include cloud agent access — Copilot Pro, Pro+, Business, Enterprise, and Max
How to try it
- Start a new task with Copilot cloud agent and choose a supported model
- Select the desired reasoning level alongside the model before delegating the task
- Consult GitHub’s documentation on choosing AI models for guidance on which reasoning level fits a given task