Comment by astrange

They have different uses. The reasoning models aren't good at multi-turn conversations.

"GPT-4.5" is the best at conversations IMO, but it's slow. It's a lot lazier than o4 though; it likes giving brief overview answers when you want specifics.