Which is better for coding: Claude or GPT?
I'm working on a Python project these days, and Claude definitely seems better at writing code, but sometimes it invents libraries that don't exist, which is bewildering... What do you all use in real-world work?
5 answers
Any source? Isn't "Claude is better at coding" ultimately just a subjective impression? It's not even a post that ran benchmarks.
I use both. Claude is definitely better at structuring code, and GPT is more convenient for debugging or explaining what caused an error. There’s no reason to stick to just one.
Making up libraries that don’t exist isn’t just a Claude problem—these days, all models do it lol. If you’re suspicious, you always have to try pip install and check the docs.
Honestly, I don't really see much of a difference. I think how you write the prompt is 90%, and the model is the other 10%. The issue of nonexistent libraries showing up also drops a lot if you add one line to the prompt: 'Only use packages that actually exist, and if you can't verify it, say so.' Blaming only the model without trying that is a bit much.
Honestly, if it's Python, the answer is to just keep both running and switch between them.