I think it depends heavily on the type of work.
For long technical documents, software architecture, repository analysis and implementation planning, Claude is still one of the tools I use most often. It is especially useful when it receives enough project context instead of just a short isolated prompt.
I would not call any model the best for every task, though. For important architecture decisions, I sometimes compare answers from different models because they often notice different risks.
Claude is worth it for me as part of a structured development workflow. It would be much less valuable if I used it only as a general chatbot.
I usually use claude max subscription and codex pro subscription. This gives best productivity and highest quality output.