Can someone explain what this toggle does for cloud Agents? Are these the models I am allowing to be used in a cloud agent run? Does the cloud agent runner limit itself to toggled on models?
Hey, thanks for the report. The “Use Multiple Models” feature works on the best-of-N principle - it’s not collaborative agents, but competition.
When you enable “Use Multiple Models”, Cursor runs the same task in parallel on different models in isolated git worktrees. They work independently - they don’t see each other and don’t collaborate. Each model races to the finish, and you pick the best result from N attempts.
This is especially useful for complex tasks where it’s unclear which model will handle it better - Composer, Claude Sonnet or GPT can give different solutions and you take the optimal one.
2 Likes
Interesting. Thanks for the info. Does this multi model toggle also work for cloud agents integrated via Github? Or does this only work if the conversation is via the Cursor IDE?
Yes, “Use Multiple Models” works the same in IDE chat and through GitHub integration with cloud agents. The functionality doesn’t differ. If you’re using cloud agents via GitHub (Linear, Slack), the toggle enables parallel task execution by multiple models - results are processed the same way.
1 Like