Google DeepMind and OpenAI have both achieved gold medal scores at the International Math Olympiad (IMO), marking a significant milestone for artificial intelligence. The experimental AI models each solved five out of six problems during the four-and-a-half-hour exam, with DeepMind scoring 35 out of 42 points.
Professor Gregor Dolinar, president of the IMO, confirmed DeepMind's result, praising the solutions as "clear, precise and easy to follow." However, only DeepMind's entry was officially recognised, as OpenAI did not formally enter the competition.
DeepMind researchers Thang Luong and Edward Lockhart stated that the reasoning capabilities demonstrated by their model would be "invaluable" for mathematicians and scientists, advancing progress towards artificial general intelligence (AGI). OpenAI researcher Alexander Wei reported his company's results online, noting that the models followed the same rules as human contestants, including no tools or internet access.
OpenAI plans to launch GPT-5 in the coming weeks, but the reasoning abilities used in the competition will not be included initially, taking "several months" to appear in public chatbots. DeepMind has not announced a release date for its Gemini Deep Think model.
