Claude vs GPT: Who’s better at coding these days?
I’ve been writing almost all my code with Claude lately, but I heard GPT has a new version out, so I’m curious. For those who’ve used both in real work, how do they feel?
Personally, Claude was easier for long refactorings, but I’ve also heard GPT was better at verifying code execution.
4 answers
I use both too, and my conclusion is "it depends on the day" lol. Even with the same prompt, the results vary so much that it's hard to say one model is definitely better.
Is there a source? I’m curious where you saw that.
Well, isn't “code execution verification” not something the model does, but rather a difference in how tools are attached? Comparing that as model performance seems a bit forced—is there even a benchmark that ran them in the same harness?
From a practical standpoint, for work that touches multiple files like refactoring, Claude definitely missed context less often. On the other hand, in the loop of running tests and fixing errors, GPT was faster at reading failure logs and pinpointing the cause. So I split my usage: Claude for design and structural changes, and GPT for debugging or writing scripts.