OpenAI’s new artificial intelligence has proven that it can perform at a human level in the world’s most prestigious mathematics competition.
OpenAI’s latest language model is capable of achieving a performance equivalent to a gold medal at the 2025 International Mathematical Olympiad (IMO). The AI model has demonstrated remarkable reasoning abilities and skill in solving complex problems.
The IMO, held since 1959, is the most prestigious mathematics competition in the world, with students from over 100 countries participating annually. Alexander Wei, a researcher at OpenAI, said about the new model: “We evaluated the model under the same conditions as actual participants — in two 4.5-hour sessions, without access to tools or the internet. Our model was able to achieve this result simply by reading the problem statements and naturally generating proofs.”
Alexander Wei further stated: “The model solved 5 out of 6 problems. Ultimately, three former IMO medalists independently reviewed the solutions, and the final score was determined by consensus: 35 out of 42 points, which is enough to earn a gold medal.”
Wei noted that OpenAI’s new AI model is still in the experimental phase and won’t be widely available for a few more months. Solving multi-hour math problems was previously impossible for AI models, but they have now reached an impressive level of understanding.
IMO problems typically require multi-page proofs, which makes training AI algorithms particularly challenging — so this achievement by OpenAI marks a significant milestone.
Before OpenAI’s recent tests, the chance of an AI earning a gold medal at the IMO was around 20%, but that figure has now reached 86%.
AI models developed by OpenAI and other companies have shown astonishing progress in programming and problem-solving, and it may not be long before they surpass human capabilities in these areas.
