Comment by o10449366

No wonder. It's performance overall was noticeably, like it had regressed to coding models from 1.5 years ago. I've try not to use claude during peak US hours because it tends to struggle more then with reasoning and correctness it seems than off hours.