OpenAI solves Paul Erdős problem that stumped mathematicians for 80 years

OpenAI says it has solved one of the most famous open problems in mathematics: Paul Erdős's unit distance problem. The problem has stumped the world's best mathematicians for 80 years, ever since it was formulated in 1946. According to the company, the solution proves that modern AI systems can carry out original mathematical reasoning and produce new proofs.

What the problem is

The problem comes from discrete geometry. It was posed by the Hungarian mathematician Paul Erdős, one of the most prolific mathematicians of the 20th century. It sounds deceptively simple: imagine many points placed on a plane, where any two points are exactly one unit apart. The question is: what is the maximum number of such points that can exist?

Despite its simple statement, the problem turned out to be one of the most stubborn open problems in mathematics. The world's best researchers tried to crack it. Every new improvement in the bound on the number of points took years of intensive research, and often required entirely new mathematical methods. Over eight decades, the problem became legendary in the mathematical community, a symbol of how a seemingly simple question can hide incredible depth.

How the model tackled the problem

OpenAI used its frontier models with advanced reasoning capabilities. The key point: the model did not simply sift through known mathematical facts, and it did not apply standard textbook methods. The system formed a new hypothesis on its own and produced an original proof, something previously considered beyond the capabilities of neural networks.

The company stresses that this is not the first time AI has shown such abilities. But each new example pushes the boundaries of what modern models can do in abstract reasoning:

• Generating original mathematical proofs
• Working with formal logic and checking the correctness of reasoning
• Developing new approaches to long-standing problems
• Independently verifying their own chains of reasoning
• Solving problems that require deep abstract analysis

Why it matters for science

Academically, it means that one of the legendary open problems, unsolved for eight decades, has finally been cracked. For AI development, the achievement shows a fundamental shift: modern neural networks can already perform meaningful mathematical reasoning that was previously considered the exclusive domain of human intelligence.

For AI researchers, it is an important signal that models are getting better and better at abstract logical thinking. That opens the door to applying AI in other complex fields that demand original mathematical thinking, from theoretical physics and cryptography to economics and biochemistry.

What it means for the future

Solving Erdős's 80-year-old problem is a milestone in AI's evolution. AI is moving beyond simply processing text patterns and known methods toward formal, original mathematical reasoning. It no longer just applies known solutions; it can generate new ideas and produce original proofs. Breakthroughs like this outline what scientists will be able to hand off to AI in research over the coming years and decades.