Interesting. Just had an instance where using gpt 4o, I got a dual response / "which is better" answer. I clearly got a o1 response (because I saw it doing the steps). Also interesting, I thought the 4o response was better, although it was confirming something, which is easy.