Post
4420
Claude Opus 4.1 is slightly better than Opus 4, but still behind GPT-5
onekq-ai/WebApp1K-models-leaderboard
onekq-ai/WebApp1K-models-leaderboard
Join the conversation
Join the community of Machine Learners and AI enthusiasts.
Sign Up
I wonder how GLM4.5 would perform, can you test them?
Claude in general has changed. Lots of complaints on the Anthropic sub. Not sure when you took your sample, but in this case it's very relevant as "he" has gone from helpful and on point, to unhelpful and error-prone:
https://old.reddit.com/r/Anthropic/comments/1njpe9c/postmortem_on_recent_model_issues/
https://old.reddit.com/r/Anthropic/comments/1nh4p53/old_claude_is_gone/
GPT-5 mini is completely stupid compared to GPT4-4o mini, and the following example is a fraction of its hallucinations and stupidity.
In this post