Google’s artificial intelligence system has achieved gold medal performance at the International Mathematical Olympiad, solving elite-level math problems that stump most humans and marking a breakthrough in machine reasoning capabilities.
The tech giant’s advanced Gemini model scored 35 out of 42 points at this year’s competition, officially earning gold medal status from IMO organizers. Only 67 of the 630 human contestants achieved the same distinction, placing the AI in the top 10 percent of participants worldwide.
“We can confirm that Google DeepMind has reached the much-desired milestone, earning 35 out of a possible 42 points — a gold medal score,” said IMO President Prof. Dr. Gregor Dolinar. “Their solutions were astonishing in many respects.”
The achievement comes as artificial intelligence companies race to demonstrate superior reasoning abilities. OpenAI announced similar results just days earlier, claiming its experimental model also solved five of six problems for an identical 35-point score. However, Google waited for official IMO certification while OpenAI tested its system independently on the same problems.
The International Mathematical Olympiad represents the pinnacle of mathematical competition for high school students, featuring problems that require hours of sustained creative thinking across algebra, geometry, number theory and combinatorics. Countries send their brightest young mathematicians to compete in what many consider the most difficult academic contest in the world.
Google’s success marks a dramatic improvement from last year, when its AI systems earned only silver medal status by solving four problems. The new system operates entirely in natural language, reading standard problem statements and producing complete mathematical proofs without requiring specialized programming languages or days of computation time.
“Our advanced Gemini model operated end-to-end in natural language, producing rigorous mathematical proofs directly from the official problem descriptions — all within the 4.5-hour competition time limit,” Google researchers explained.
This represents a significant departure from previous AI mathematical achievements, which typically required narrow, task-specific programming. The IMO-capable models are built on the same general-purpose language systems used for everyday tasks, suggesting broader implications for artificial intelligence development.
The rapid progress has impressed even veteran mathematicians. Prof. Sir Timothy Gowers, an IMO gold medalist and Fields Medal winner who evaluated AI solutions, called the results “very impressive, and well beyond what I thought was state of the art.”
Both Google and OpenAI emphasized that their achievements should complement rather than replace human mathematical talent. The companies noted that IMO competitions exist primarily to inspire young people to pursue mathematics careers, not to pit machines against humans.
“Our leap from silver to gold medal-standard in just one year shows a remarkable pace of progress in AI,” Google noted, while acknowledging the continued importance of nurturing human mathematical talent.
The timing reflects intense competition among AI companies to demonstrate reasoning capabilities that approach human-level performance. However, both firms indicated their gold medal systems remain experimental and won’t be available to the public for several months.
As AI systems increasingly match human performance in complex cognitive tasks, the mathematical olympiad achievement signals another milestone toward artificial general intelligence. The ability to solve problems requiring genuine mathematical insight and creativity suggests these systems may soon become valuable tools for professional mathematicians and scientists.
Yet the achievement also raises questions about the pace of AI development and its implications for education and human expertise. When machines can match the performance of elite teenage mathematicians, the landscape of human-AI collaboration in scientific fields may be fundamentally shifting.
The competition was held this year in Australia, where human contestants spent two sessions of 4.5 hours each tackling six exceptionally difficult problems. The AI systems operated under the same time constraints, producing detailed mathematical proofs without internet access or external tools.
IMO organizers found Google’s solutions “clear, precise and most of them easy to follow,” suggesting the AI had not only solved the problems correctly but presented its reasoning in ways that human mathematicians could readily understand and verify. This combination of accuracy and clarity represents a significant advance in making AI mathematical reasoning both powerful and accessible.