Comment by icelancer

I think the point is that a larger model with far less reasoning time solves the problem just fine. Which of course is the tradeoff: The smaller the model, the more reasoning you need to get decent answers to tough questions.