Will AI solve 0 more Millennium Prize Problems in 2026?

- [2026-09-08] OpenAI announced its internal AI model (10,000 coordinated agents, ~88 hours) produced a proof that OpenAI published a claimed solution on Tuesday to the Navier-Stokes existence and smoothness problem, one of seven Millennium Prize Problems, and the proof was produced by an internal AI model, not human mathematicians. This is explicitly **excluded** from the qualifying problems in this market's rules. https://qz.com/openai-ai-navier-stokes-millennium-prize-math-090826 - [2026-09-08] OpenAI stated it does not intend to claim the Clay $1M prize; OpenAI stated in its announcement that it would not claim the $1 million Clay Millennium Prize for the solution it posted, and as of September 2026, the Clay Mathematics Institute currently considers the problem "active." https://en.wikipedia.org/wiki/Navier–Stokes_existence_and_smoothness - [2026-09-11] OpenAI told the New York Times it has already made significant progress on another one of the problems—but didn't specify which one. This is unconfirmed as to which of the five qualifying problems it concerns. https://www.explainx.ai/blog/millennium-prize-problems-ai-solved-fact-check-2026 - [2026-09-18] A fact-check update notes rumors later named the unspecified second problem as Hodge, but this remains **unverified**: Hodge and Birch–Swinnerton-Dyer now have unverified rumors of near-solves, plus OpenAI's confirmed-but-unnamed claim of substantial progress on a second Millennium problem after Navier-Stokes — but no named paper, Lean artifact, or lab confirmation on either conjecture. Until a primary source names the problem, keep reading those rows as rumor, not progress. https://www.explainx.ai/blog/millennium-prize-problems-ai-solved-fact-check-2026 - [2026-08-11] Anthropic (via an unreleased Claude research model) made progress on a Riemann-Hypothesis-adjacent bound, not the hypothesis itself: Anthropic announced that an as-yet-unreleased model had made significant progress on the Riemann hypothesis, significantly increasing the lower bound of solutions for which the hypothesis holds true, but contemporary AI models still can't solve it. Anthropic's own post states this result shows that AI models like Claude can extend the impact and reach of mathematicians' ideas in new and sometimes surprising ways, even though it couldn't resolve the Riemann hypothesis itself. https://www.anthropic.com/research/riemann-zeta - [2026-09-26] Fact-check tracker's most recent update (dated today) reports Anthropic's newest math result is in Yang-Mills-adjacent amplitude territory, not a mass-gap proof: "Anthropic reported a planar N=4 SYM six-gluon MHV amplitude at nine loops validated by Lance Dixon's SLAC group — amplitude technology in the same physics neighborhood, not a mass-gap proof. The Clay mass-gap question itself still has no AI proof attempt as of this update." https://www.explainx.ai/blog/millennium-prize-problems-ai-solved-fact-check-2026 - [2026-09-26] Same tracker states, as of its latest update, that "None of the six currently unsolved Millennium Prize Problems has been certified solved by the Clay Mathematics Institute." and no lab has publicly named a solved qualifying problem. https://www.explainx.ai/blog/millennium-prize-problems-ai-solved-fact-check-2026 - [2026-09-10] Wikipedia's Millennium Prize Problems page confirms only Poincaré is officially solved, and records OpenAI's Navier-Stokes claim as a "proposed counterexample," not a certified Millennium solution: "the only Millennium Prize problem to have been officially declared solved is the Poincaré conjecture... In September 2026, the American artificial intelligence company OpenAI presented a proposed counterexample to the Navier–Stokes existence and smoothness problem, saying that they do not intend to claim the Millennium Prize for the result." https://en.wikipedia.org/wiki/Millennium_Prize_Problems - [2026-09-24] Polymarket's companion multi-outcome market ("How many more Millennium Prize Problems will AI solve in 2026?") shows trader consensus heavily favoring zero additional qualifying solutions: outcome "0" priced at 57-58%, "1" at 32-33%, "2" at 6-7%, higher counts near-zero. Market context notes: "Trader consensus around 0 or 1 additional solution in the remaining months of 2026 reflects the remaining problems' extreme complexity, verification timelines for formal proofs, and the substantial resources needed to replicate such results quickly." https://polymarket.com/event/how-many-more-millennium-prize-problems-will-ai-solve-in-2026 - [2026-09-13] Commentary (Zvi Mowshowitz) flags the pace-of-progress narrative around this event, noting: "As recently as April this year, prediction markets gave AI less than a 40% chance of solving any Millennium Prize Problem before 2030." This underscores how fast sentiment has shifted after the Navier-Stokes claim, even though Navier-Stokes itself doesn't count for this market. https://thezvi.wordpress.com/2026/09/13/brand-new-ai-solves-a-millennium-prize/ - [2026-09-08 ongoing] Independent scrutiny of the Navier-Stokes claim continues (priority dispute with Buckmaster/Alpöge, and questions over whether the "forced" variant matches the real Clay problem): "On 8 September 2026, artificial intelligence company OpenAI claimed a solution to the Navier–Stokes existence and smoothness problem with a smooth external force. However, both results have gaps with the original Navier–Stokes problem, which excludes Euler equations and requires the external force to be smooth." This dispute is separate from the market's qualifying problems but illustrates the high scrutiny bar any future claim would also face. https://en.wikipedia.org/wiki/Navier%E2%80%93Stokes_priority_controversy - [2026-09-11] Related Polymarket market "Which Millennium Prize Problem will AI solve next?" currently shows Hodge Conjecture as favorite at 53%, indicating market expectation is concentrated on a future (not-yet-occurred) event rather than a confirmed one. https://polymarket.com/event/which-millennium-prize-problem-will-ai-solve-next BOTTOM LINE: As of September 26, 2026, no eligible AI lab has made a qualifying announcement (solution claim) for any of the five listed problems (BSD, Hodge, P vs NP, Riemann, Yang-Mills); the only AI Millennium-adjacent claim to date (OpenAI's Navier-Stokes result) is explicitly excluded by the rules, and all other developments (Anthropic's Riemann bound, Yang-Mills-adjacent amplitude work, rumored-but-unnamed "second problem" progress) are progress/partial results, not solution announcements — supporting the "0 more solved" outcome, consistent with the ~57-58% market-implied probability. NOT FOUND: No official Clay Mathematics Institute statement declaring any of the five qualifying problems solved; no named, lab-confirmed solution announcement for Hodge, BSD, P vs NP, or Yang-Mills as of this report's date; no evidence of a scheduled lab announcement before December 31, 2026 for any of the five qualifying problems.

- [2010-03-18] Historical base rate: of the seven original Millennium Prize Problems, exactly one has ever been solved in 25 years since the prize was announced — the only Millennium Prize problem to have been officially declared solved is the Poincaré conjecture, and the Clay Institute awarded the monetary prize to Grigori Perelman in 2010. This gives a ~25-year base rate of one solved problem, i.e., a very low historical annual "solve rate" for the class of problems generally. - [2003] The one historical solve (Poincaré) took roughly a century of cumulative human effort building on Thurston and Hamilton's work, and even then the conjecture was declared proven in 2006 after four years of debate and three 300-page-long publications filling in details of the proof — illustrating that verification/consensus after an announcement has historically taken years, not weeks. - [2026-09-08] The reference class shifted abruptly this year: OpenAI announced on 8 September that it has generated a solution to the Navier–Stokes equations, showing that they can break down, claiming to have cracked one of the Millennium Problems. However, this specific Polymarket question explicitly excludes Navier-Stokes from the qualifying list, so this landmark event does NOT count toward this market's "Yes" resolution. - [2026-09-09] Confirming the exclusion is intentional and central: only one Millennium Prize problem has been officially solved so far, the Poincaré Conjecture; OpenAI said the company does not intend to claim the Millennium Prize for the Navier-Stokes result. - [2026-08] Prior to the Navier-Stokes claim, expert/market sentiment on AI solving any Millennium Problem was low: as recently as April 2026, prediction markets gave AI less than a 40% chance of solving any Millennium Prize Problem before 2030. - [2026-08-13] The closest AI has come on any of the five *qualifying* problems for this specific market is incremental: Anthropic does not expect that the techniques Claude used will lead to proving the Riemann hypothesis, but its work serves as the latest example of the speed of progress in AI models' mathematical capabilities. - [2026-09-19] A related but distinct Polymarket/Kalshi contract on "AI lab announces another Millennium Prize solution" (which does count the 5 qualifying problems, same list as this question) has been extremely volatile on rumor rather than fact: Polymarket's "which problem next" ladder had Hodge at 43 to 45 cents, Birch and Swinnerton-Dyer at 26 to 27, Yang-Mills in the mid-teens, Riemann at three cents, and P versus NP at about a cent, while "no problem solved by any lab by the end of 2027" traded at 12 to 13. - [2026-09-19] Kalshi's equivalent "OpenAI announces another Millennium Prize solution before 2027" contract swung wildly on unconfirmed rumors: from 28¢ to 84¢ and back down, driven by a Hodge Conjecture rumor and a vague OpenAI statement to the New York Times, with no actual qualifying announcement made. The report notes "OpenAI has not announced a second solution. What the week's tape ... did to the price." (paraphrased from source, and per the article's own framing that nothing has settled the contract). - [2026-09-24] Directly on-point: the actual Polymarket market matching this question ("How many more Millennium Prize Problems will AI solve in 2026?") currently prices "0" (i.e., this question's "Yes") at 57-58%, "1" at 33%, "2" at 6-7%, and higher counts at low single digits, per the live order book fetched from Polymarket. - [2026-09] Expert/mathematician sentiment is skeptical of near-term further solves: Terence Tao said "the dynamic is now that of frenetic competition" and that "the indiscriminate use of AI is turning the subject into a meaningless production quota 'game' that ultimately is of very little benefit, either to mathematics or to the world." A declaration by 25 Fields medalists reportedly stated that AI companies' goals and the mathematical community's goals are "severely misaligned" (per the OddsShopper summary of Tao's blog post), reinforcing that credible verification/endorsement of any full solution is likely to be slow and contested even if a lab claims one. - [2026-09] An outside factual-check source concluded that no genuine AI progress exists yet specifically on the harder qualifying problems: P vs NP, Hodge, and Yang-Mills — problems with no known "almost there" partial result to extend — remain completely untouched, while Riemann, which had 112 years of incremental human bounds to build on, got a real jump. BOTTOM LINE: The pure historical base rate for solving any Millennium Prize Problem is extremely low (one confirmed solve in 25 years, taking years for consensus), but 2026 has seen a paradigm-shifting event (OpenAI's Navier-Stokes claim) that does not count for this market's five qualifying problems, and current market pricing (Polymarket ~57-58% on "0" additional solves, i.e., this question's Yes) and expert sentiment both suggest further qualifying solves within the remaining ~3 months of 2026 are possible but more likely than not to NOT occur, given no qualifying announcement has yet been made on Riemann, P vs NP, Hodge, BSD, or Yang-Mills. NOT FOUND: No specific historical base-rate dataset exists for "AI solving math conjectures on a fixed annual cadence" beyond the current 2026 episode (this being the first year AI has plausibly solved any Millennium-class problem at all), so pre-2026 base rates are drawn entirely from the human-only track record (1 in 7 over 25 years). Could not find a stable, non-rumor-driven Metaculus community forecast specifically calibrated to the Dec 31, 2026 deadline for these five problems.

**Most likely path to YES: OpenAI names and announces the Hodge Conjecture (or another qualifying problem) as solved before Dec 31, 2026, with the announcement structured to satisfy the market's "presented as the lab's own work" clause.** - [2026-09-11] OpenAI's own statement to reporters is a primary-source admission — not rumor — that it has made progress beyond Navier-Stokes: OpenAI has said that it has made "substantial progress" on another Millennium Prize Problem , confirmed independently to the New York Times and Washington Post per a later summary of the reporting chain. - [2026-09-17] The Information (paywalled but corroborated by multiple outlets), citing a source inside OpenAI, reports the specific target and imminence: not long after OpenAI swept in to announce it had solved the famous Navier-Stokes math problem, the company is allegedly homing in on another Millennium Prize Problem, with employees at the AI firm expecting to soon crack the Hodge Conjecture. - [2026-09-17] The same reporting frames this as a deliberate, near-term release being managed for PR reasons rather than a research bottleneck: "It could take the company longer to announce the solution, though, because it's trying to figure out how to collaborate with the math community to make the announcement without triggering another PR nightmare," The Information paraphrased its source as saying. This suggests the blocker is messaging strategy, not unfinished mathematics — a mechanism that could resolve quickly once OpenAI settles its communications approach. - [2026-09-21] OpenAI's own blog post (not rumor) states a dramatically compressed capability timeline for the same model line: an internal model it began training on August 28 has, beyond the Navier-Stokes Millennium Prize problem announced two weeks earlier, now resolved more than 100 long-standing open problems across most areas of mathematics... The pace of its progress in mathematics has surprised the mathematicians within OpenAI. - [2026-09-21] Crucially, multiple outlets reporting on this Sept 21 disclosure state the 100+ list is claimed to already include a second Millennium Problem: Among the solved problems is reportedly a second Millennium Prize Problem, the Hodge conjecture. If accurate, the solve already exists internally and only an eligible-lab announcement (which the market requires) remains outstanding — a much lower bar than solving the problem from scratch. - [2026-09-21] OpenAI's formation of an outside oversight body signals it is actively building the "responsible disclosure" scaffolding needed to make a credible, qualifying announcement rather than retreating from further claims: OpenAI has announced the formation of an independent advisory group on mathematics and artificial intelligence, in collaboration with mathematicians and hosted by the Institute for Advanced Study at Princeton... Initial members include François Charles, Camillo De Lellis, Timothy Gowers, Martin Hairer, Nikhil Srivastava, Ulrike Tillmann, Ravi Vakil, Edward Witten, and Melanie Matchett Wood. This body exists specifically to shepherd exactly this kind of release, making a managed near-term Hodge announcement more plausible than an indefinite delay. - [2026-09-21] Notably, OpenAI explicitly carved the "pace of internal progress" out of the advisory group's remit: OpenAI has excluded the pace of its research from the group's advisory role. This indicates OpenAI retains full unilateral control over the timing of any further Millennium-adjacent announcement, i.e., no external gatekeeper can force a delay past 2026. - [2026-09-14] Grassroots sentiment among close OpenAI-watchers already anticipated an imminent named announcement, e.g., one widely viewed X post stated flatly: This week or next OpenAI is going to announce another millennium problem they have solved, the hodge conjecture. While unofficial, this reflects how compressed the expected timeline is among people tracking OpenAI's internal signals, only 3.5 months before the Dec 31 deadline. - [2026-09-08] The Navier-Stokes precedent establishes the mechanism and speed at which OpenAI can go from internal result to public claim: OpenAI's own post states the goal is to report on the substantial progress of our AI models... This milestone represents substantial work by mathematicians and AI researchers, and Quanta reported the entire Navier-Stokes proof process, from generation to formal Lean verification, took roughly a group of 10,000 autonomous AI agents... running on an advanced model not available to the public just 88 hours — showing the lab is both willing and technically able to move from result to public claim in days, not years. - [2026-09-24] Even skeptical secondary analysis frames the Hodge claim as the "narrowest, most authoritative" fact in the chain rather than pure rumor, distinguishing OpenAI's own words from anonymous amplification: OpenAI says "substantial progress" on an unnamed Millennium Prize problem... "Substantial progress" on a second, unnamed Millennium Prize problem. No conjecture identified. This is the narrowest, most authoritative claim in the chain. This is a company-confirmed fact, not merely a social-media leak, and it is the single fact standing closest to a qualifying announcement. BOTTOM LINE: The strongest YES path is OpenAI converting its company-confirmed "substantial progress on a second Millennium Prize problem" (Sept 11) plus reported internal completion of a Hodge solve within its 100+-problem haul (Sept 21) into a formal, lab-attributed announcement before Dec 31, 2026; the prerequisites — an already-existing internal result, a purpose-built external advisory apparatus to manage disclosure, and OpenAI's retained control over release timing — are largely already in place, leaving primarily a PR/messaging decision rather than a mathematical one as the remaining barrier. NOT FOUND: No official OpenAI statement naming Hodge (or any qualifying problem) as solved; no Lean/paper artifact for Hodge; no confirmed publication date for the "next" announcement.

- **[2026-09-11]** The Clay Mathematics Institute itself confirmed via president Martin Bridson that even the (excluded) Navier-Stokes claim would be evaluated **"deliberately unhurried"** and **"absolutely rigorous,"** and that CMI's own rules require a solution must be published in a suitable academic venue, then at least two years must pass, and it must receive general recognition from the mathematical community before it is considered officially solved , per the CMI's official rules page, meaning even a "solved" qualifying announcement this market counts (lab claim) would face a multi-year human-verification backlog before any institutional confirmation — a bar this market's resolution criteria bypass, but which conditions how cautiously labs may phrase future claims. (https://seoulinside.substack.com/p/did-openai-just-solve-navierstokes) - **[2018 rules, still in force]** Per CMI's official rules: Before CMI will consider a proposed solution, all three of the following conditions must be satisfied: (i) the proposed solution must be published in a Qualifying Outlet (see §6), and (ii) at least two years must have passed since publication, and (iii) the proposed solution must have received general acceptance — illustrating the institutional culture of extreme caution that any lab claiming a Hodge/BSD/Riemann/P-vs-NP/Yang-Mills solve would be up against, reinforcing why labs may hold back a "solved" framing before Dec 31, 2026. (https://www.claymath.org/millennium-problems/rules/) - **[2026-09-17/22]** The only reported "second problem" lead (Hodge) remains anonymously sourced: There is no public OpenAI paper, named theorem, second Lean formalization, or Clay Mathematics Institute certification , and The accurate headline is therefore: OpenAI says it is close to another Millennium problem, and Hodge is the strongest public lead — not a solve. (https://kingy.ai/news/openai-next-millennium-prize-problem-hodge/) - **[2026-09-22]** Even the most detailed tracing of the rumor chain concludes: The Information reports, sourced anonymously, that OpenAI is close to solving the Hodge Conjecture, with people familiar and staff expecting a solve "relatively soon." ... No paper, no code, and no statement from the Clay Mathematics Institute is publicly available on the Hodge Conjecture . (https://fourweekmba.com/ai-openai-hodge-conjecture-clay-scoreboard/) - **[2026-09-21]** After the Navier-Stokes credit controversy, OpenAI itself signaled caution rather than acceleration: reporting indicates Any announcement could be delayed, though. After the PR crisis around Navier-Stokes, OpenAI wants to get the messaging right this time — direct evidence of self-imposed slowdown on any second qualifying-problem announcement, cutting against a rapid "Yes" resolution. - **[2026-09-08/10]** The Navier-Stokes episode itself demonstrates how contested even a completed, Lean-verified AI proof can become before any "solved" framing sticks: a citation dispute arose in which at least 2 of them (such as the sphere packing problem) had been criticized for failing to properly cite prior work — for example, Yeshiva University's Stephen Miller argued that the sphere packing proof reused the argument of his 2016 paper without citation... Some mathematicians have even used the phrase "academic misconduct" for this pattern , and a priority dispute with Buckmaster/Alpöge remains unresolved as of mid-September — a template for the scrutiny any future BSD/Hodge/Riemann/Yang-Mills/P-vs-NP claim would immediately face, likely delaying an unambiguous "the lab claims this is solved" statement. - **[2026-09-11]** The 25 Fields Medalists' declaration the goals of the AI companies and the goals of the mathematical community are severely misaligned... the push by AI companies to solve mathematical problems as a benchmark is detrimental to the science of mathematics, and to the mathematical community , and by Sept 16 had at the time of posting has over 7000 signatures... By 16 September, the declaration's website listed more than 7,200 signatories — creating reputational/community pressure that may make labs more cautious about issuing unambiguous "solved" claims (as opposed to "progress") on the remaining qualifying problems, since qualifying language requires presenting the solution as the lab's own work rather than partial results. - **[2026-09-12]** Even for the excluded Navier-Stokes result, outside experts state verification is incomplete: Johns Hopkins mathematical physicist Gregory Eyink also held that no one on the human side has yet fully verified this proof , and Three days later, on September 11, the Clay Institute moved a step further, stating that the Navier–Stokes existence and smoothness problem "could have been settled" and that this now triggers its formal, multi-year verification process — signaling that "solved," even when a lab claims it, is treated with institutional hedging that could bleed into how any future qualifying-problem announcement is phrased (i.e., as "progress" rather than an unambiguous solve, which would not satisfy this market's resolution criteria). - **[2026-09-26]** As of the most recent tracker update available, two of these seven problems have been at the center of real AI news in the last six weeks, and one of those stories is a cautionary tale about believing the headline before checking the proof , with P vs NP, Hodge, and Yang-Mills still having no named AI proof attempt — confirming that three of the five qualifying problems have zero lab-claimed progress at all as of the report date. - **[2026-09-24]** Market pricing directly on this question (companion Polymarket contract) shows the "0 additional solves" outcome (this question's Yes) trading at 57-58%, moderately favored but far from certain — reflecting genuine two-sided uncertainty rather than settled consensus, and leaving ~3 months for a qualifying announcement that has not yet materialized on any of the five listed problems. BOTTOM LINE: No eligible lab has issued an unambiguous "solved" announcement on any of the five qualifying problems as of Sept 26, 2026; the sole concrete Millennium-adjacent claim (Navier-Stokes) is explicitly excluded, faces an unresolved priority/citation dispute, and even that claim triggered only a multi-year CMI verification process rather than instant certification — a pattern of institutional caution, community backlash (Fields Medalists' declaration), and OpenAI's own signaled slowdown that together make it structurally more likely labs continue issuing "progress" rather than "solved" statements through year-end, favoring NO on further solves (i.e., "Yes" on this question of zero additional solves). NOT FOUND: No primary-source (lab blog, press release, or arXiv/Lean artifact) naming Hodge, BSD, Riemann, P vs NP, or Yang-Mills as solved by any eligible lab; no scheduled announcement date from any lab for a second qualifying-problem claim; no CMI statement on any of the five qualifying problems beyond its Navier-Stokes commentary (which doesn't count for this market).