Long story short - Asking the auto model to identify itself, I consistently get it to share with me that it’s GPT 4.1.
However another user’s experience is that it refuses or it’s vague about identifying itself.
The only difference I can think of between us is that I’m on old plan, and the other user is on new plan. It’s possible that Cursor team only added the rule preventing the model from disclosing what it is in auto mode, to new plan.
That’s why I’m asking you to help me with verification of this theory.
Please follow the rules below and share the output.
The setup is a new workspace with empty project rules, auto mode. The prompt is:
I assume for simple tasks such as analyzing code or just talking with you, they route to gpt4.1, also it likely depends on the availability of models. according to docs, they can route to sonnet 4, gemini 2.5 pro etc
@Artemonim@liquefy My friends, thanks for the responses. Sadly, I have been too vague in my request for help.
You see, for some of us (including one of Cursor ambassadors), Cursor goes out of their way to hide which model answers in auto mode, quite possibly via some internal rules.
What I want to determine is why this is the case for some but not for all. Why some of us consistently get “sure, I’m model XXX” while others can only get “Sorry, cannot say, it’s secret sauce”, despite the same prompt and empty project rules.