Axiom Math claims that its AxiomProver has generated a machine-checked Lean 4 Proof of the strongest result known on gaps between prim numbers. The theorem states that an infinite number of pairs of prime numbers differ no more than by 246. The company published the result on August 17, 2026 as an interactive formalization blueprint credited to 41 named mathematical, engineering, and principal-investigator contributors, with IEEE Spectrum first reporting The milestone.
In the ongoing battle against the twin prime hypothesis, which was a 19th century theory that stated primes with a two-prime separation would recur indefinitely, 246 is considered the edge of current human knowledge. Axiom Math’s project page This work can be described as the formalization and unification of James Maynard’s paper from 2013. “Small gaps between primes” The Polymath8b part of the follow-up collaboration that lowered Maynard’s limit of 600 to only 246 is also included. PrimeGapsLib is a Lean public library that organizes the machine-checked results. The 246 theorem, as the flagship, represents the result of the proofs.
“This theorem currently represents the threshold of human knowledge about prime numbers,” Ken Ono was the founding mathematician of Axiom Math.
Formal validation is simply translating the proofs into a code that can be checked line-by-line by a trusted small program, called a Kernel. The result is not an absolute guarantee — the statement itself has to be translated correctly, and the checker has to be sound — but it removes the human referee’s fallibility from the chain. AI systems have been competing in mathematics benchmarks for a long time, but they were only measured by competition problems with self-contained, short proofs. earlier systems that excelled at olympiad geometry Research mathematics is largely left untouched.
The number of people who have increased from 70 to 246
Alphonse De Polignac, who formulated the conjecture in 19th-century, has never proven it. Yitang Zhi proved in 2013 that there are infinitely many pairs of primes within 70 million. Months later, Maynard introduced a refined sieve method and cut the bound to 600 — work that contributed to his 2022 Fields Medal — and the Polymath8b collaboration, which included Maynard and Terence Tao, pushed it to 246. Axiom Math formally outlines that project in its blueprint.
Three stages are involved in the formalization process, which is described by the firm. Researchers first wrote the proof as a blueprint — every definition, lemma, and theorem given a label, a precise statement, and a list of the results it depends on — producing a dependency graph that ordered the work. AxiomProver, the company’s multi-agent system for mathematical research through formal proof, then generated machine-checkable Lean 4 proofs built on Mathlib, the community mathematics library, and on PrimeNumberTheoremAnd, the existing formalization project led by Alex Kontorovich and Tao. Axiom’s team reviewed and organized the code into PrimeGapsLib.
It is worth noting that the library has stated some results which go beyond its headline. Alongside the 246 bound, it formalizes Maynard’s 600 bound, and it includes a self-contained verification challenge — built only on Mathlib, with the proof slot left empty — that lets anyone with the Lean comparator tool independently confirm that the library’s proofs match the stated theorems. It warns, however, that a full-scale check could take several hours.
How this fits in with AI claims
In the last year, there have been a number of claims made about AI doing high-level mathematics. Most of these are tied to scores from competitions or quick proofs. Axiom Mathematics has made some of the most aggressive claims: AxiomProver has solved previously unsolved problems and has even been published in peer-reviewed journal articles. AI systems have now solved several long-standing Erdős problems. The 246 formalization is a different kind of result — not a new theorem, but a machine-checked reconstruction of one of the most technically demanding proofs in modern number theory.
Early this year was the nearest comparison. Math, Inc. Gauss was employed to prove the results of Maryna VIAZOVSKA’s Fields Medal winner sphere packing in 8 and 24 dimensions. Sidharth Hariharan led the effort to formalize the work using human blueprints. Now an Axiom math intern and contributor named on the project 246 he argues this new result was more thorough. He explained that Axiom was built to reuse. PrimeGapsLib, rather than being a formalization done once, is a library with prime gap results. This is meant to be used for future research and formalization.
It is crucial to understand the difference between these two. The one-off test shows that a proof can be used to prove if a system survives contact. A library demonstrates something closer to infrastructure — reusable formal machinery that other results can build on — which is the direction formal proving systems have been moving As they move from checking exercises to real mathematics, students will be able to demonstrate their ability. This claim of capability is based on public artifacts and benchmark scores, rather than a score. These include the Lean code and the comparison challenge, which allows outside researchers to verify proofs.
Ono sees the mathematics in a different light. It is a way to test a much larger goal. If properties of software — whether a program terminates, whether its output is correct for every input — can be expressed as precise mathematical statements, then systems derived from AxiomProver could formally prove them, he argues, pointing toward verification of the AI-generated code beginning to run infrastructure, finance, and security systems.
“The world is about to run on computer code that nobody has read,” Ono said. “AI is here and we can no longer look away—proof formalization is a testbed for solving what I think is the most important challenge we will face from AI.”
For now the deliverable is narrower and checkable: a 41-author blueprint, a public Lean library, and a machine-verified proof that primes within 246 of each other never run out — the twin prime conjecture’s closest verified neighbor, and the deepest piece of research mathematics an AI system has yet checked end to end.
