Are you noticing a difference in code quality between Claude and GPT these days?
Recently I had both do refactoring work, and the difference I felt was quite significant. GPT feels like it quickly produces a draft, while Claude feels like it thinks about the structure for a long time? Of course, it depends on the prompt.
I'm curious which one everyone uses more in actual work.
3 answers
I also feel Claude is better at refactoring. Especially when making changes across multiple files, GPT often misses inconsistencies between different parts. In the end, I think the answer is to use both and cross-check them.
Honestly, these days all the models seem pretty similar... I feel like talking about worrying over the architecture is just dressing up the fact that the output is slow. If there were benchmark numbers, maybe, but generalizing based only on subjective experience is a bit questionable.