An AI model from Google reached gold-level scores at an international mathematical competition in Australia that ended Sunday, though it still didn’t beat some human competitors who represent the world’s elite math talent younger than 20 years old. The gold marks — given to roughly the top 10% of student contestants — were the highest the AI models have ever scored, indicating the technology is catching up to human genius. OpenAI also said a few days earlier one of its models had achieved a similar score on the problems, but without officially participating in the competition.
Google’s advanced version of Gemini and OpenAI’s experimental reasoning model both solved five of six math problems, which cover algebra, geometry, number theory, and combinatorics (a counting math), the representatives of the companies said. Five human contestants received perfect scores.