I’ve used both Auto mode and Composer 2.5, but both seem to consume my plan quota.
Auto chooses the “best model for the task”, optimizing for speed and intelligence (but mostly for price, if we’re being honest).
As such, this is usually a worse model than the ones you can use via API…
Anyways, Cursor now has 2 quotas, API usage and auto usage. Composer models are included in auto.
Hey, thanks for the request. This is expected behavior, not a bug.
Auto mode is where Cursor picks the best model for the task, balancing speed and quality. On individual plans, there are two separate included usage pools:
- Auto + Composer: both Auto and Composer 2.5 use this pool, so they share the same quota.
- API: a separate pool for third-party models like Claude, GPT, Gemini, etc.
So what you’re seeing is by design. Auto and Composer 2.5 both draw from the same Auto + Composer pool. You can check the usage breakdown anytime at cursor.com/dashboard under Usage.
More details about models and pools: Models & Pricing | Cursor Docs
Let me know if you’ve got any other questions.