Model Selection for Claude and Codex Agents
Claude and Codex agents now let you pick models on github.com. Choose Anthropic or OpenAI models when running tasks—included with Copilot.
TL;DR
- Claude and Codex third-party agents now support model selection on github.com
- Claude users can pick from Sonnet 4.6, Opus 4.6, Sonnet 4.5, or Opus 4.5
- Codex users get GPT-5.2-Codex, GPT-5.3-Codex, or GPT-5.4
- Included with existing Copilot subscriptions; admins must enable policies
New
- Model selection for Claude agent — Choose between Sonnet 4.6, Opus 4.6, Sonnet 4.5, or Opus 4.5 when running tasks, giving you access to the latest Anthropic models as they ship.
- Model selection for Codex agent — Pick from GPT-5.2-Codex, GPT-5.3-Codex, or GPT-5.4 to match your task complexity and latency needs.
What You Need to Do
- Model selection works like the Copilot cloud agent — pick your model when kicking off a task.
- Copilot Business/Enterprise — Your admin must enable the Anthropic Claude or OpenAI Codex policy in enterprise settings.
- Repository owners — Enable the agent from Settings > Copilot > Cloud agent.
- Access is included with your existing Copilot subscription; no additional cost.
See the GitHub docs on third-party agents for full details.
Source: GitHub Changelog
