I used Cursor and Chat GPT 5 last night for the first time. Before I could even ask Chat GPT 5 about my issue it had scanned the .cpp file in question (because it was open in the editor) and had discovered some possible issues, one of which was the issue in the code. I confirmed that and gave it more description of the error behavior. It identified the problem in the code, and suggested two different CORRECT solutions (one simple, one more complex but "perfect"). I opted for the simple one. It did it. One tiny problem remained, I pointed it out, it fixed it.
This was much better than Gemini or CoPilot on the exact same issue and the exact same commit pointer in my repo. Both of them suggested the same wrong solution and got themselves further and further wrong as they went.
I have something that both Gemini (via GCA) and CoPilot (Claude) analyzed and came up withe the same diagnosis. Each of them made the exact same wrong solution, and when I pointed that out, got further wrong.
I haven't tried Chat GPT on it yet, hoping to do so soon.
This was much better than Gemini or CoPilot on the exact same issue and the exact same commit pointer in my repo. Both of them suggested the same wrong solution and got themselves further and further wrong as they went.
So, I guess as of today, Chat GPT 5 leads. YMMV