Key Points
- Google and OpenAI’s AI models both achieved “gold medal” scores at a top international math competition.
- This marks the first time AI has reached this level, representing a significant breakthrough in its reasoning ability.
- The AIs solved the problems using natural language, not just specialized code. Experts believe AI could soon be used to help solve major problems.
- Google utilized its Gemini Deep Think model, whereas OpenAI employed a new, computationally intensive method.
In a breakthrough for artificial intelligence, Google and OpenAI have announced that their AI models achieved “gold medal” scores at the International Mathematical Olympiad (IMO), a prestigious global math competition for high school students. This is the first time any AI has reached this level of performance.
The achievement signals a significant leap in AI’s reasoning abilities, bringing the technology closer to rivaling human intelligence. Both companies’ models solved five out of six complex problems. Crucially, they did this using general-purpose AI that processed the problems in natural language—much like a human would—rather than relying on specialized, math-only code.
While Google officially collaborated with the IMO to have its results certified, OpenAI had its model’s performance independently graded by former IMO medalists. Experts believe this success means AI is less than a year away from being able to assist human mathematicians in tackling major unsolved problems.
OpenAI achieved its result with a new, “very expensive” method that allowed the model to “think” for longer and explore multiple lines of reasoning simultaneously. Google used its Gemini Deep Think model, which solved the problems within the official 4.5-hour time limit.
There was some minor drama over the timing of the announcements, as Google waited for the official release date while OpenAI published its results a couple of days earlier, saying it had received permission to do so.
