OpenAI Model Cracks 80-Year-Old Erdős Math Problem

An OpenAI reasoning model has cracked open a mathematical puzzle that has stood for nearly eight decades, and leading mathematicians are calling it a milestone for artificial intelligence in pure research. The result, announced this week and validated by independent experts, disproves a long-held belief tied to a famous question first posed by Hungarian mathematician Paul Erdős in 1946.

The puzzle is called the planar unit distance problem. It sounds deceptively simple: if you place a set of dots on a sheet of paper, how many pairs of those dots can be exactly the same distance apart? Erdős conjectured that the number grows only slightly faster than the number of dots themselves. For 80 years, mathematicians have searched for arrangements that pushed the limit higher, and the best examples — based on neat square grids — seemed to suggest Erdős was right.

The OpenAI model showed otherwise.

A New Family of Arrangements

"For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids," OpenAI wrote in announcing the result. "An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better."

The model did not produce a complete answer to the broader question of exactly how fast the number of equal-distance pairs can grow. What it did was prove that the established upper limit was too low, by drawing on ideas from different branches of mathematics to construct arrangements that beat the grid. In other words: the ceiling people thought existed turned out not to be the ceiling.

Critically, the company says the work was done not by a system custom-built for math, but by a general-purpose reasoning model that breaks problems into smaller steps and explores them. The result has been written up in a companion paper involving human mathematicians.

Validation From Skeptics

OpenAI has been here before, and not always well. A claimed Erdős breakthrough last year turned out to be the model regurgitating results it had absorbed from existing literature, embarrassing the company.

This time, the work has been examined and endorsed by some of the same people who criticized the earlier claim. Thomas Bloom, the mathematician who maintains the canonical Erdős problems website, co-authored the companion paper. He wrote that the AI reached its result by "persevering down paths that a human may have dismissed as not worth their time to explore."

Fields Medalist Tim Gowers, also writing in the companion paper, described the result as "a milestone in AI mathematics."

What This Means for AI and Research

Mathematicians are quick to note that humans were still essential. "While the original proof produced by AI was completely valid, it was significantly improved by the human researchers at OpenAI and the many other mathematicians involved in the present paper," Bloom wrote. "The human still plays a vital role in discussing, digesting and improving this proof, and exploring its consequences."

The collaboration model — AI generates candidate ideas and proof sketches, humans refine and validate — is starting to look like a viable workflow for problems that have resisted progress for decades.

Andrew Rogoyski of the Institute for People-Centred AI at the University of Surrey called the announcement evidence that AI is reshaping how humans approach hard problems. "It's becoming clear that AI is impacting the world of creative thought and will become a fundamental tool of future scientific research," he said.

Why It Matters Beyond Math

The planar unit distance problem belongs to a field called discrete geometry, which sounds abstract but underpins everything from coding theory to crystal physics to the design of computer networks. Pushing the boundary of what arrangements are possible has practical echoes in disciplines that depend on packing, signaling, and structure.

More broadly, the achievement suggests that AI is moving past mimicry and beginning to genuinely contribute to discovery — not by guessing answers, but by patiently trying ideas that humans would have written off. For an 80-year-old problem that survived everything mathematicians threw at it, the breakthrough is a quiet but significant turning of the page.