OpenAI says its AI cracked a maths problem unsolved since 1946 without human help

OpenAI says one of its latest AI reasoning models has solved a long-standing mathematics problem that researchers had been working on for decades. The company claims the model independently discovered a new mathematical proof linked to the “planar unit distance problem,” a geometry problem first proposed by legendary mathematician Paul Erdős back in 1946. Also Read: CapCut x Gemini Integration: Gemini now wants to edit your videos too

The announcement has quickly grabbed attention across the AI and research communities because this time, OpenAI says the model did not simply retrieve an existing solution from research papers or online databases. Instead, the company claims the AI came up with a completely new approach on its own. Also Read: Gemini Omni to AI-powered Search: Everything Google announced at I/O 2026

This also comes months after earlier criticism around claims involving GPT-5 allegedly solving unsolved maths problems, which later turned out to be based on already existing literature.

What exactly is the maths problem?

The planar unit distance problem may sound complicated at first, but the basic idea is easier to understand.

The problem looks at how points can be arranged on a flat surface while keeping as many pairs of points exactly one unit apart from each other as possible. For years, mathematicians believed that grid-like arrangements were among the best possible solutions.

Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946.

For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids.

An OpenAI model has now disproved that… pic.twitter.com/j2g3Ze0zEG

— OpenAI (@OpenAI) May 20, 2026

According to OpenAI, its AI model discovered a new family of constructions that performs better than the older assumptions researchers had relied on for decades.

The company says external mathematicians later reviewed the proof and supported the findings. Reports mention names like Noga Alon, Melanie Wood, and Thomas Bloom, who examined the result after the announcement.

Why researchers are paying attention

What makes the development important is not just the maths itself, but how the result was achieved.

OpenAI says the system used here was a general-purpose reasoning model instead of a specialised tool designed only for mathematics. The company claims the model was able to maintain long chains of reasoning and connect different mathematical ideas together without human intervention.

That is very different from the usual image most people have of AI systems, which are mostly known for generating text, images, or writing code.

Solving mathematical proofs is far more demanding because even a tiny mistake can break the entire argument. Researchers usually spend months or even years checking proofs carefully before accepting them.

The result is already being discussed as a possible milestone for AI reasoning systems because it suggests these models may slowly move beyond assisting researchers and start contributing original ideas of their own.

Impact may go beyond mathematics

The implications could stretch far beyond geometry problems.

OpenAI believes systems capable of handling long and difficult reasoning tasks may eventually help researchers in fields like biology, physics, engineering, materials science, and medicine.

Many scientific problems involve analysing huge amounts of information while connecting ideas across different domains. That is where companies working on AI believe reasoning-focused models could become useful in the future.

Add Techlusive as a Preferred Source

At the same time, researchers are still approaching such claims carefully, especially after previous AI breakthroughs were later questioned or corrected.