OpenAI claims to have solved an 80-year-old mathematical problem, for real this time

OpenAI claims that its new inference model has produced an original mathematical proof that refutes a famous unsolved conjecture in geometry, first proposed by Paul Erdös in 1946.

If this sounds familiar to you, it’s because this isn’t the first time OpenAI has made such a bold claim. Seven months ago, Kevin Weil, a former vice president at the AI ​​giant, posted on X: “GPT-5 has found solutions to 10 (!) previously unsolved Erdős problems and has made progress on 11 more.”

It turns out that GPT-5 didn’t actually solve those problems; I just found solutions that already existed in the literature.

Taunts from rivals such as Yann LeCun and Google DeepMind CEO Demis Hassabis followed, and Weil soon resigned from his position prematurely. Today, at least, it appears that OpenAI has not made the same mistake twice. Along with the announcement, the company published accompanying notes supporting the rebuttal by mathematicians such as Noga Alon, Melanie Wood, and Thomas Bloom, who runs the Erdös Problems website, previously calling Weil’s post a “sensational distortion.”

“Nearly 80 years ago, mathematicians believed that the best possible solutions looked roughly like square grids,” OpenAI posted on X. “OpenAI’s model has now refuted that belief, discovering an entirely new family of top-performing constructions.”

The company said this represents “the first time that artificial intelligence has independently solved a prominent open problem central to the field of mathematics.” The proof, according to OpenAI, came from a new general-purpose reasoning model, not a system specifically designed to solve mathematical problems or even this problem in particular.

OpenAI says this is important because it means that AI systems are now better able to piece together long, difficult chains of reasoning and connect ideas across domains in ways that researchers may not have explored before. This has implications for biology, physics, engineering and medicine.

“AI helps us more fully explore the cathedral of mathematics we have built over centuries,” Bloom said in a statement. “What other unseen wonders await us?”