AI systems make new gains in mathematical research

Advanced AI models are gaining traction as research assistants in mathematics, with recent breakthroughs suggesting they can help tackle problems that have challenged experts for decades. OpenAI’s GPT-5.6 Sol reportedly provided the key construction behind a new paper disproving the 150-year-old Maxwell Conjecture, while human researchers completed the formal proof and credited the model for the central insight.

OpenAI’s experimental research systems have also been reported to solve multiple long-standing problems in mathematics and theoretical computer science using relatively modest computing resources. The results point to a shift from AI as a calculation tool toward AI as a collaborator capable of proposing promising routes for expert review.

China’s RedNote reported another milestone with dots-note-3.0, which became the first AI system to achieve a perfect score of 42/42 on the International Mathematical Olympiad. The competition requires full mathematical proofs reviewed line by line by human judges, making the result a notable benchmark for reasoning rather than simple answer generation.

The momentum reinforces the view that mathematics may be one of AI’s strongest domains because proofs offer clear verification, rapid feedback, and structured reasoning. Human experts remain essential, but AI-generated insights are already shaping work across mathematics, physics, and theoretical computer science.