Every time I ask it to correct something in the code or make a suggestion how to improve something in the file, instead of highlighting a relevant part of the code, it repeats the entire file in the answer, almost always hitting the maximum char response limit.
Moreover, it’s rarely highlighting that relevant part like GPT-4 normally does with comments.
You know, like
var = false // set this to false
that sort of thing.
Does anybody experience something like this? Is this model specific issue or Cursor?
+1. Hope this is fixed. I went back to basic GPT-4 because of this, even though its a lot more money. (Im using openrouter api but it sounds like this problem is universal).
I agree with it writing the entire code file for one line of correction. This happens with visual studio copilot ALOT! So much so that I did have to move over to cursor and vs code
I actually noticed lately this is not just 4o, but regular GPT-4 model.
I’m explicitly asking it to only show relevant parts of the code (multiple times, lol), and yet it copy-pastes the entire file without highlighting anything even with comments, inevitably hitting the limit: