An unannounced Anthropic AI model has taken on one of the biggest unsolved problems in mathematics

The AI industry is increasingly grappling with the question of whether these systems can genuinely contribute to advancing scientific knowledge, rather than simply automating administrative tasks. A recent example, reported by TechCrunch, shows that an experimental model developed by Anthropic—one that hasn't yet been released to the public—has made notable progress on one of the toughest problems in mathematics: the Riemann hypothesis.

First formulated in 1859 by German mathematician Bernhard Riemann, this conjecture is considered one of the most important unsolved problems in mathematics, with deep implications for number theory and the distribution of prime numbers. The Clay Mathematics Institute has listed it among its seven "Millennium Prize Problems," offering a $1 million reward for a complete solution.

What the Anthropic model actually achieved

According to TechCrunch, the model did not solve the Riemann hypothesis, nor did it produce a complete proof. What impressed researchers was the system's ability to generate mathematical reasoning paths considered genuinely valuable, going beyond simply reproducing knowledge already present in its training data.

This kind of progress matters to the AI business community because it demonstrates a shift from models that carry out predefined tasks toward systems capable of exploring open problems with no known solution. The distinction is crucial: solving math exercises with a known answer is a very different feat from contributing to a problem that thousands of mathematicians have worked on unsuccessfully for decades.

Why this matters for the AI market

The announcement comes amid fierce competition among major AI companies, which are investing heavily in models capable of advanced reasoning. Anthropic, backed financially by Amazon and Google, is competing directly with OpenAI and Google DeepMind to demonstrate the superiority of its systems' cognitive capabilities.

Progress on unsolved mathematical problems isn't just academically significant. It also serves as a strong commercial argument for companies looking to sell AI models as valuable tools for scientific research, financial engineering, or cryptography—fields where number theory plays a central role.

For now, Anthropic has not confirmed a release date for this model, and full technical details about the methodology used have not been made entirely public, according to the cited source.