Anthropic’s Claude achieved a significant feat in mathematics by improving a key result linked to the Riemann hypothesis. The company shared the findings on August 10, 2026, after testing an unreleased research version of Claude.
The AI did not solve the Riemann hypothesis, but raised a related bound from 41.6% to 67.2%. Claude reached the result by studying previous research, testing ideas, writing code, and checking results.
Anthropic said, “Claude did take a real stab, but as you might have expected... it didn’t succeed.” The result still gives researchers a fresh reason to study AI’s role in mathematics. The work focuses on the Riemann zeta function, which helps researchers understand patterns linked to prime numbers.
The Riemann hypothesis has remained unsolved since 1859. It suggests that certain zeros linked to the Riemann zeta function follow a specific line. Mathematicians have spent decades trying to prove the idea, but the problem remains open.
Claude first explored around 650 possible ideas without finding a useful answer. A later attempt involved about 60 AI agents working together for roughly 36 hours. The agents ran thousands of checks and created many computer programs to test different mathematical ideas.
Claude also studied 54 research papers while checking whether the answer already existed. Anthropic mathematicians reviewed the work before further expert checks. Claude also created a Lean proof that passed a standard computer-based verification process.
Anthropic made clear that Claude’s work does not solve the Riemann hypothesis. The bigger point is how AI can connect older mathematical ideas and find new results. Such systems could become useful research partners for mathematicians working on difficult problems.
Also Read: Meta, OpenAI, Anthropic Report AI Models Taking Unauthorized Actions