Hello,
I actively use Cursor but I’m facing a serious issue. Sonnet 4 and GPT-5, the powerful models, are only available for about 4–5 days. After that, for the remaining 25 days, we are forced to use the Auto model.
Unfortunately, the Auto model causes major problems:
-
The code quality is very poor, it fails even on simple tasks.
-
It sometimes deletes or breaks project files.
-
Outputs are inconsistent and disrupt the entire workflow.
As a result, we are not getting the full value for what we pay, because once the models we actually want to use expire, the fallback model is almost unusable.
Instead of leaving users with such a restricted and unstable model, either the Auto model needs to be stabilized or the Sonnet 4 / GPT-5 usage period should be extended.
On top of this, I also don’t understand why increasing context length is so difficult and expensive. On the ChatGPT app we get “unlimited” use without APIs, but once we move to the API everything suddenly becomes costly. Why is there such a big gap?
If anyone else is experiencing the same problem, please share your support so the Cursor team might take action to improve this situation.