I find it hard to reconcile Anthropic’s surging revenues and market cap with my personal experience that Opus 4.6 is visibly dumber and less useful than 5.4, and the fact that when I’m constrained to using it by the circumstances of the task, I am demonstrably less productive. Perhaps I’m just using and evaluating Claude for a different context, or maybe I’m genuinely noticing a capabilities difference. If it’s the latter though, I don’t expect Anthropic to be able to hold on for long, no matter what their revenues are, as OpenAI’s models continue to aid their R&D.
I find it hard to reconcile Anthropic’s surging revenues and market cap with my personal experience that Opus 4.6 is visibly dumber and less useful than 5.4, and the fact that when I’m constrained to using it by the circumstances of the task, I am demonstrably less productive. Perhaps I’m just using and evaluating Claude for a different context, or maybe I’m genuinely noticing a capabilities difference. If it’s the latter though, I don’t expect Anthropic to be able to hold on for long, no matter what their revenues are, as OpenAI’s models continue to aid their R&D.
With the release of 5.5, I continue to be confused about this. They are robustly ahead on benchmarks.