Add Dynamic Model Switching in GitHub Copilot CLI During Runtime

[iwangbowen]

Currently, users can only change the model in GitHub Copilot CLI by setting the COPILOT_MODEL environment variable before starting the CLI, which requires restarting the application for changes to take effect. This limits flexibility, especially during interactive sessions where switching models on the fly could be useful for testing or adapting to different tasks. Implementing a command or option to dynamically switch models (e.g., via a CLI flag like --model gpt-5 or an interactive command) would enhance usability without consuming additional premium requests. This feature would align with the existing quota system and improve the overall developer experience by allowing seamless model adjustments mid-session.

[Michspirit99]

Adding to this, we should be able to do this in application during runtime. OpenCode, Claude Code, Gemini CLI, Cursor, all do this.