Beginning in the mid-2020s, large language models, including reasoning models, have made increasing progress in mathematical proofs. Models from OpenAI and Anthropic, in particular, have achieved a number of significant results with minimal human intervention.
List In May 2026, OpenAI announced that an internal model had produced a disproof of the unit distance problem. In July 2026, mathematician Levent Alpöge presented an explicit counterexample to Jacobian conjecture, which he found with Claude Fable 5. In August 2026, OpenAI announced ten results achieved by an internal version of Astra, an unreleased model, including a construction establishing the existence of non-sofic groups. However, a supposed disproof of Connes's rigidity conjecture was later refuted. In August 2026, Anthropic announced that an unreleased version of Claude increased the proven share of Riemann zeta function zeros on the critical line from 41.6% to 67.2%. In August 2026, Levent Alpöge and Ava Howell used Claude to find the two elliptic curves with the largest known rank (at least 30 and 31, respectively). In August 2026, Alpöge claimed to have used Claude to prove that the 6-sphere admits a complex structure. The result has not yet been independently verified. If confirmed, this would resolve the 78-year old Hopf problem.
See also Computer-assisted proof
References
