How many more Millennium Prize Problems will AI solve in 2026?

The market’s Yes side asks whether AI solves zero more qualifying Millennium Prize Problems in 2026, and the balance of evidence still favors that outcome. As of October 10, the only major 2026 claim involving a Millennium-level problem is OpenAI’s claim on Navier-Stokes, which is explicitly excluded by the market rules. For the five qualifying problems that remain in scope, there has been impressive progress and some partial-results reporting, but no clearly announced, complete solution from an eligible AI lab. That matters because the resolution standard is strict: the lab must publicly announce that the problem has been solved, and partial advances do not count.

The strongest reason to expect Yes is simple timing and historical precedent. These problems are extraordinarily difficult, and even with rapid AI-assisted theorem discovery, a fully accepted or at least unequivocally announced solution to one of them is a very high bar. The latest OpenAI release reportedly contained many substantial mathematical results and some work connected to the Riemann hypothesis, but it did not claim a full proof of any qualifying problem. If the most advanced and visible AI lab in this space has not crossed that threshold by early October, the remaining window is short. Late-year announcements are possible, but they would need to be both real and framed as complete solutions by an eligible lab, which is a much narrower condition than producing progress.

Market pricing is consistent with that view. The Yes contract at 0.775 implies the market already expects roughly a three-in-four chance of zero additional qualifying solutions, which is a strong signal that participants think the base rate remains very low. The residual risk is not trivial, though. One plausible upside scenario is that an AI lab already has a proof or near-proof for one of Riemann, Hodge, Birch and Swinnerton-Dyer, P versus NP, or Yang-Mills and chooses to announce it before year-end. If that happens, the market would flip immediately to No. But absent a concrete public claim, the most defensible forecast is that AI’s 2026 breakthroughs stop short of the official threshold for any qualifying Millennium Prize Problem.