OpenAI's Internal Model Solves Over 100 Math Problems in One Month

OpenAI's internal model has successfully solved over 100 long-standing math problems after just one month of training, which began on August 28, 2026. Among these, the model reportedly addressed two Millennium Prize Problems, including the Hodge conjecture. This achievement was highlighted by Jakub Pachocki, Chief Scientist at OpenAI, who noted the rapid progress made by the model.

Timothy Gowers, a Fields Medalist and member of the Advisory Group on Mathematics and Artificial Intelligence based at the Institute for Advanced Study, expressed concerns that AI-generated solutions could undermine conceptual understanding among mathematicians. He emphasized the need for effective communication about the value of human mathematical expertise in light of these advancements.

OpenAI has clarified that the advisory group will not influence the pace of its internal progress in mathematics. Gowers also mentioned that large language models capable of tackling significant math problems will be publicly available within months. This follows OpenAI's recent establishment of an advisory group focused on mathematics and AI, which aims to address the implications of these developments.