Bottom line: OpenAI has used its next-generation language model to disprove an 80-year-old mathematical problem—for less than $1,000 in computational costs. A milestone in AI-driven mathematics.
OpenAI has used its next generation of large language model, presumed to be GPT 5.6, to disprove an 80-year-old mathematical conjecture. The Erdős Planar Unit Distance Problem has been solved—and for less than $1,000 in computational costs.
OpenAI is celebrating a major scientific breakthrough: the next model in its series (presumed to be GPT 5.6) has succeeded in disproving the legendary Erdős Planar Unit Distance Problem. This mathematical problem was posed by renowned mathematician Paul Erdős approximately 80 years ago and long stood as one of the most difficult unsolved puzzles in combinatorial geometry.
What is remarkable about this success is not only the scientific achievement itself, but also the efficiency: the solution was reached with computational resources costing less than $1,000. This demonstrates the growing capacity of modern AI systems to tackle complex mathematical problems.
The focus of the tech industry is currently also on other major developments. The planned IPO filing by SpaceX and related companies will be treated separately and should be appropriately recognized on the actual IPO date.