Google, OpenAI AI Models Win Gold at International Math Olympiad

In 2025, AI models developed by Google DeepMind and OpenAI reached a historic milestone by earning gold medals at the International Mathematical Olympiad (IMO), marking the first time AI systems scored at the top level in this prestigious high school math competition. Both models solved five of the six problems using general-purpose reasoning models that processed mathematical concepts in natural language rather than relying on symbolic computation. Google’s AI, based on its Gemini Deep Think model, completed the exam under official conditions alongside human competitors, while OpenAI achieved its score independently without official IMO participation, generating some controversy over timing and verification. The AI success was attributed to innovations like scaling test-time compute and reinforcement learning with curated mathematical solutions, enabling multiple parallel reasoning paths and longer processing times. Experts, including Brown University professor Junehyuk Jung, foresee this breakthrough enabling future collaborations between AI and mathematicians to tackle unsolved research problems and impact fields beyond mathematics, such as physics. These developments highlight AI's growing potential in advanced reasoning and problem-solving, bringing it closer to assisting in real-world scientific and academic challenges.