Claude was not the best model for reasoning even vs 4o, and it's quite visible once you start giving it more complex logical puzzles. People seem to like it more mostly because the way it speaks is less forced and robotic, and it's better at creative writing usually, but if you need actual _intelligence_, GPT is still quite a bit ahead of everybody else.

Now I don't think that it's because OpenAI has some kind of secret sauce. It rather seems that it's mostly due to their first mover advantage and access to immense hardware resources thanks to their Microsoft partnership. Nevertheless, whatever the reason their models are superior, that superiority is quantifiable in money.