OpenAI's experimental model wins gold at the International Mathematical Olympiad

An experimental OpenAI model has won gold at the International Mathematical Olympiad, solving 5 out of 6 problems and scoring 35 out of 42, OpenAI researcher Alexander Wei said. The result matches the performance of the world's best math students and demonstrates a significant breakthrough in AI's ability to reason logically, Engadget reports.

The model, which has not yet been officially unveiled, participated in a simulation of the Olympiad problems without human intervention. It not only found the correct solutions, but also formulated full mathematical proofs. OpenAI called it a "long-awaited breakthrough in reasoning" that the company has been developing for several years.

The International Mathematical Olympiad is one of the most prestigious competitions for students, attracting over 600 participants from dozens of countries each year. In 2025, only 67 of them won gold — about 10% of the total. That's the result of OpenAI's experimental AI.

Despite the achievements, the company does not plan to publicly release the model in the coming months. According to Wei, it is not yet integrated into consumer products.

OpenAI previously aimed to create models capable of abstract thinking and reasoned analysis, and the International Mathematical Olympiad was seen as one of the benchmark tests for creative and logical abilities.

The new achievement confirms that AI is becoming increasingly good at automating not only routine tasks, but also tasks traditionally considered the preserve of humans.