GitHub Copilot cloud agent adds customizable reasoning levels
GitHub Copilot cloud agent now allows users to select a reasoning level for AI models that support it, enabling control over response complexity and token consumption. This feature is available on paid Copilot plans and aims to improve answers for complex problems. Users can pick a reasoning level alongside the model when initiating a task.
- →Customize reasoning level for Copilot cloud agent
- →Reasoning levels impact cost and response quality
Features (1) ›
- Customize reasoning level for Copilot cloud agent
Users can now set a reasoning level for GitHub Copilot cloud agent tasks, allowing control over model output complexity and token usage. This feature helps tailor responses for intricate problems and is available on various paid Copilot plans.
Notes (1) ›
- Reasoning levels impact cost and response quality
Higher reasoning levels can yield better answers for complex issues but consume more tokens and credits. Users select this level at the start of a task alongside the AI model.
https://github.blog/changelog/2026-08-03-customize-the-reasoning-level-for-copilot-cloud-agent
Related releases
- GitHub Copilot introduces Agent Plugins 1.0 for cross-client development GitHub Changelog ·
- GitHub Details Organization-Level Repository Rule Insights GitHub Changelog ·
- GitHub Enterprise Server 3.22 Release Candidate Now Available GitHub Changelog ·
- GitHub Copilot for JetBrains adds persistent memory, Ollama integration, and enterprise controls GitHub Changelog ·
- Automatically Migrate Branch Protection Rules to Repository Rulesets GitHub Changelog ·
- GitHub Copilot's MAI-Code-1-Flash model slated for deprecation GitHub Changelog ·