OpenAI Model Solves 80-Year-Old Math Problem That Stumped Experts

An artificial intelligence model built by OpenAI has disproved a famous math conjecture that stumped researchers for nearly 80 years. The Erdős unit distance conjecture, first posed by legendary mathematician Paul Erdős in 1946, asked a deceptively simple question about geometry that no human could fully answer — until now.

What the AI Actually Did

The conjecture deals with unit distances in discrete geometry. In plain terms, it asks: given a set of points on a flat surface, what is the maximum number of pairs that can sit exactly one unit apart? For decades, mathematicians believed the answer grew only slightly faster than the number of points itself. OpenAI's reasoning model — a type of large language model, or LLM, designed for step-by-step logical thinking — proved that belief wrong. It discovered an infinite family of point configurations where the count of unit-distance pairs grows at a rate of n raised to the power of one plus a small number called delta. Princeton mathematician Will Sawin later refined that delta value to 0.014, confirming the result independently.

Why This Matters for AI and Science

What makes this breakthrough stand out is that the model was not purpose-built for mathematics. It is a general-purpose reasoning system that happened to crack the problem during a broader evaluation on open math challenges. Fields Medalist Tim Gowers, one of the most respected mathematicians alive, reviewed the proof and said he would recommend it for publication without any hesitation. That endorsement is a first for a proof generated entirely by artificial intelligence. The achievement suggests that modern AI systems are moving beyond pattern recognition into genuine logical reasoning, a shift that could accelerate discoveries across science, engineering, and medicine.

This milestone arrives at a pivotal moment for OpenAI, which is reportedly preparing a confidential filing for an initial public offering. A research win of this scale strengthens the company's position as a leader not just in chatbots and productivity tools but in fundamental scientific discovery. If AI can disprove conjectures that baffled human experts for eight decades, the question is no longer whether machines can do real science but how fast they will reshape it.