Different models for chat, apply, and Cmd-K

Different models for different purposes would be a nice important feature! Easy to implement - a more granular “model selection” would offer this.

At the moment, I would choose models like this:

  • Opus - Chat (and “new AI project”, if it’s still there, good chance Opus can do it)
  • GPT-4 - “apply changes” (since Opus fails here)
  • Haiku - Cmd-K (quick, cheap, good enough)

Switching models with every step is not feasible. I reverted back to GPT-4, since I always “chat” and “apply” in succession.

I think I’d use Cmd-K, if it were quicker (Haiku).

This would also dramatically lower your costs, dear Cursor crew :slight_smile:


Was wondering the same thing - this would be massively helpful, to aim the right speed / reasoning ability to the right type of task.

1 Like

+1 :fire:
A necessary addition for this is choosing which API key to use for which model and enabling the possibility to use them in an intertwined manner based on which model is chosen. So basic GPT-4 and Claude 3 Opus (10 per day) through the use of the Pro subscription, while also being allowed when pressing
“Command+/” to use models through the API key. Right now it’s really annoying to switch between on-off API key depending on the task.


+1 :fire: :fire:

1 Like

+1 :slight_smile:

1 Like

@truell20 Sorry to bother you, but I’d like to know if this feature is on the roadmap, including possibly the other QoL improvements, stated in my previous post.
Two times today I wanted to quickly do a CMD+K in file while Opus was still generating in Chat, which caused the reply from Opus to stop and be removed from the chat. The more I use Cursor, the faster I become, the more I run into this problem. Thank you in advance.