Last week, we evaluated o3 and o4-mini on FrontierMath on the “high” reasoning setting and found that o4-mini set a new state-of-the-art!
We’ve now also run both models on low and medium reasoning. How well do they score, and how much did this cost? Results in thread!
