I have auto turned on for the AI model, but it seems to be ALWAYS using gpt 5 in “thinking” mode. How do I turn the thinking off?
1 Like
You can’t, the best answer I got was “turn off auto and choose a model” Funny, cause when I got an issue with how the tokens are calculated I get advised to “turn on auto”
Cursor in its current state is completely useless, the fastest response you get is like 30 seconds at best…
Does the thinking mode output count towards any usage any limits?
ah interesting, so that means that they sometimes use GPT-5 for the Auto mode? Until now most of the Auto mode responses were, when I asked it to present itself, Sonnet 4 (non-thinking). I guess they’re trying it out, but yeah GPT-5 is reaaally slow compared to Sonnet 4!
Im finding that it’s ONLY using GPT5 for auto mode.
Not only. its a set of premium models.
1 Like
the lack of transperancy on auto-mode makes me VERY sceptical that it uses any premium models at all. It for sure used GPT5 during the release weekend, somehow it was even worse than the old auto-mode model which seems to be mostly an old version of gpt4.
For what it’s worth i’m very happy with automode recently and appreciate the cursor team for using new models like gpt5 in auto. thx cursor team 
1 Like
As a workaround just add this to your user rules “At the end of your work plz add a simple line about which model was used, when auto was selected.” and it will give you the model most of the time.