Copilot coding agent is our asynchronous, autonomous background agent. Delegate a task to Copilot, and Copilot works in the background in its own development environment in the cloud. Once Copilot finishes work, it opens a pull request and requests your review.

In December, we launched model selection for Copilot Pro and Pro+ users. Today, we’ve opened up the model picker to Copilot Business and Copilot Enterprise users, so you can choose the model that’s right for your task.

When assigning an issue to Copilot on github.com, or when starting a task from the agents tab, agents panel, GitHub Mobile, or the Raycast launcher, you can use the dropdown to pick from the following models:

  • Auto
  • Claude Opus 4.5
  • Claude Opus 4.6
  • Claude Sonnet 4.5
  • Claude Sonnet 4.6
  • GPT-5.1-Codex-Max
  • GPT-5.2-Codex
  • GPT-5.3-Codex

If you leave the setting on Auto, Copilot optimizes speed and performance by automatically choosing a model based on availability.

If your administrator has not enabled any of the models listed above, the model picker will not be shown, and Claude Sonnet 4.6 will be used automatically.

To learn more, head to our documentation about model selection in Copilot coding agent.