Everyone raves about Opus, but for coding, I’ve found Sonnet seems to be better. Is this because Opus is better at the creative thinking and problem solving, but when it comes to writing code based on a good, narrow specification, Sonnet is just as good, if not better?
Or am I completely misunderstanding?
I find that I get better results from Sonnet than Opus. What I’ve learned is that there is no ‘Best’ model for everything, they all handle problems a bit differently. I flip-flop back and forth between Gemini and Sonnet, depending on the nature of the problem.
The only real difference I’ve found with Opus is cost and speed. It takes longer, and the result seems to be the same as Sonnet. Its impossible for me to measure though so its just the ‘vibe’ of the thing. It could also be that I’m not pushing the models hard enough for Opus to shine 
I find OPUS to be better at narrow specification and doing changes that align more with the requirement, and following your codebase patterns even better than you specified. Narrow is relative though so not sure we are considering the same thing
I have for 2 days used OPUS and it is not bad at all. At least as good as Sonnet, and somewhat better. I see larger changes at once. Read benchmarks. Anthropic claims OPUS is better.
have you tried opus 4 on cursor or claude code?