Google DeepMind Achieves Official Gold at IMO 2025, Scores 35 Points – OpenAI Also Hits Gold-Level
Google DeepMind model result was officially graded and certified by IMO
The International Mathematical Olympiad (IMO) is a mathematical competition organized for high school students. It was first held in Romania in 1959 with seven participating countries. Today, the IMO has grown to include over 100 participating countries. Each countries are represented by six elite individuals who compete to solve six difficult problems in algebra, combinatorics, geometry, and number theory.
Last year, with the merging of Alphaproof and AlphaGeometry 2, Google DeepMind's system was able to solve four out of six problems at the IMO with an awarded score of 28 points. This year, Google Gemini Deepthink model earned 35 points for solving five of six IMO problems correctly. Google's Gemini Deepthink model was awarded gold-medal performance.
After the 2025 IMO, an OpenAI model attempted the IMO 2025 problem. While complying with the rules applicable to human contestants, OpenAI model solved five out of six problems. Three formal IMO medalist independently graded the proof solution submitted by OpenAI model. After unanimous consensus, OpenAI earned 35 points in total, which is liable for gold medal-level performance
Comments