Comment by do_not_redeem

A third party tried this experiment with publicly available models. OpenAI did half as well as Gemini, and none of the models even got bronze.

https://matharena.ai/imo/