I recently switched over to anthropic models, and they seem to have a better “attitude” compared to openai.

I also like the (marginally) fewer hallucinations, but it means more refusals (when uncertain) in API calls. Looking forward to trying out this “step change” when it’s released.