OpenAI Models Find Proofs for Three Scientific Problems

On October 26, 2026, OpenAI announced that its frontier models obtained a proof of the local smoothing hypothesis for all dimensions, a refutation of the Exponential Time Hypothesis (ETH), and rigorous proofs of the existence of quantum chromodynamics and pure Yang–Mills theory in four dimensions, publishing the manuscripts as 'raw research material' for independent review by external experts without NDAs.

Compiled by Sergey KostenchukPublished 2026-10-26Updated 2026-10-27

2026-10-26 Research OpenAI

🧠 OpenAI Models Find Proofs for Three Scientific Problems

On October 26, OpenAI reported three results from its frontier models: the local smoothing hypothesis was proven, a refutation of ETH was constructed, and proofs of the existence of QCD and Yang–Mills theory were claimed. Manuscripts are released without NDAs as 'raw material'; key results are formalized in Lean.

🌍 If Impagliazzo, Paturi, and Williams confirm the ETH refutation, it undermines hundreds of conditional results in graphs, biology, and databases. OpenAI is building an AI science pipeline: raw manuscripts, Lean, and independent review.

👤 Preprints and proofs are available at github.com/openai/math. One should not believe that 'AI solved a millennium problem': the existence of Yang–Mills is only part of the Clay problem, and the physics has not been published.

Source 1: https://openai.com/index/building-scientific-understanding-with-academic-communities/ Source 2: https://www.unite.ai/openai-says-its-models-produced-three-proofs-in-math-cs-and-physics/

Sources