I was referring to the o1 AIME accuracy figure (x log scale compute, y is % (not log)) and similar https://openai.com/index/learning-to-reason-with-llms/