Is cursor serving a different model than requested?

Long Story Short, been running out of claude requests. Started incorporating Auto more and was curious what model it was serving up, so I added a cursor rule telling the chat to say what model it is at the end of each response.

This came up when I was using o3 earlier for a more demanding task. It would seem I got 4o served instead of o3.

It was still saying 4o when I started a new chat with o3 (although old chat was completely using o3 as well).

The picture that says claude 3.5 through o3 platform, I have no idea what that means. I was using sonnet 4 in the chat earlier so I assumed it was some hallucination but after seeing it also report 4o later when I requested o3, I’m skeptical.

I’m hoping it’s just something to do with cursor’s system prompts or something else, otherwise, it’s a really bad look for cursor given the recent usage debacles.

But, I’m on 1 year free student pro. Tabs autocompletes have been amazing and since everything is free, I don’t really have anything to complain about, everything been pretty good.

Am curious if anyone else have similar cases.



  • OpenAI models don’t know themselfs
  • Claude and Gemini don’t know excact version
  • Grok know
1 Like

You can also provide a Request ID and ask Dev team to check it.