Which math problems will AI solve in 2026?
Each market in this event asks whether an eligible AI lab will announce that it has solved one named math problem by the end of 2026. The key rule is that the lab must claim the solution as its own work; an announcement that only credits outside researchers who used the lab's tools does not count.
Leading right now: Riemann Hypothesis 76%, Ramsey Number R(5,5) 20%, Strong Goldbach Conjecture 18%, Twin Prime Conjecture 16%.
| Market | Chance of yes | Deadline |
|---|---|---|
| Will AI solve the Riemann Hypothesis in 2026? | 75.5% | Jan 1, 2027, 4:59 UTC |
| Will AI determine the Ramsey number R(5,5) in 2026? | 20.5% | Jan 1, 2027, 4:59 UTC |
| Will AI solve the strong Goldbach conjecture in 2026? | 17.5% | Jan 1, 2027, 4:59 UTC |
| Will AI solve the twin prime conjecture in 2026? | 15.5% | Jan 1, 2027, 4:59 UTC |
| Will AI solve the Hadamard matrix conjecture in 2026? | 14.5% | Jan 1, 2027, 4:59 UTC |
| Will AI solve Beal's conjecture in 2026? | 13.5% | Jan 1, 2027, 4:59 UTC |
| Will AI solve Legendre's conjecture in 2026? | 12.8% | Jan 1, 2027, 4:59 UTC |
| Will AI solve the odd perfect number problem in 2026? | 10.9% | Jan 1, 2027, 4:59 UTC |
| Will AI solve the Hodge Conjecture in 2026? | 9.5% | Jan 1, 2027, 4:59 UTC |
| Will AI solve the Birch and Swinnerton-Dyer Conjecture in 2026? | 9.0% | Jan 1, 2027, 4:59 UTC |
| Will AI solve the Collatz conjecture in 2026? | 8.0% | Jan 1, 2027, 4:59 UTC |
| Will AI solve the Yang-Mills existence and mass gap problem in 2026? | 7.5% | Jan 1, 2027, 4:59 UTC |
| Will AI solve P versus NP in 2026? | 7.3% | Jan 1, 2027, 4:59 UTC |
| Will AI solve the Hadwiger-Nelson problem in 2026? | 6.5% | Jan 1, 2027, 4:59 UTC |
Odds read from Polymarket at 22:18 UTC on Nov 21, 2026. $198K traded so far, $87K in the last 24 hours.
How this market resolves
Yes: An eligible AI lab, or an official representative of it, announces by December 31, 2026, 11:59 PM ET that the specific problem named in that market's title has been solved. The announcement must present the solution as the work of the lab, its researchers or its models, alone or jointly with outside researchers. It must claim a full solution, not partial progress.
No: No qualifying announcement is made by the deadline. This includes the case where only partial results or progress are announced, where an announcement only credits or congratulates outside researchers, or where the claim comes from a company or person who is not an eligible lab.
When it really ends
December 31, 2026, 11:59 PM ET (Eastern Time). In December ET is UTC-5, so this is 2027-01-01 04:59 UTC. The endDate field in the data (2027-01-01T04:59:00Z) matches the rules for every listed market.
What counts
- An announcement by a named lab such as Anthropic, Google, OpenAI, Meta, Microsoft, DeepSeek, Harmonic, Axiom Math or Math Inc, or by their subsidiaries and research divisions.
- An announcement by any other company that has a model on the MMLU-Pro leaderboard at the time of the announcement.
- A statement by an official representative of an eligible lab.
- A solution credited to the lab, its researchers or its models, including joint work with outside researchers.
- A consensus of credible reporting, which the rules say may also be used as a resolution source.
What does not count
- Partial results or progress toward a solution.
- An announcement that only credits or congratulates outside researchers, even if they used the lab's models, compute or research credits.
- Announcements by companies not on the list and not on the MMLU-Pro leaderboard.
- Announcements by individuals or organizations that are not eligible labs, including universities.
- A claim made after the deadline.
Fine print people miss
- Confirmation from the Clay Mathematics Institute or any other body is not required. The lab's own announcement is enough.
- The rules do not say who decides whether an announcement is a full solution, or how a disputed claim is treated. The rules do not require peer review or independent verification.
- The rules do not define what counts as a solution for each specific problem, for example whether a proof or a counterexample is needed.
- Whether a problem is 'solved' for open-ended items such as the Hadamard matrix conjecture or Hadwiger-Nelson is left to how the announcement is worded and reported.
- The eligible list is long, and any company with a model on the MMLU-Pro leaderboard at announcement time also qualifies, so the pool of eligible announcers is wide.
- Each problem is a separate market with its own title, and the same rules apply to each.
What moves the price
- A lab publishing a paper or post that claims a full proof of a named problem would push that market toward Yes.
- Credible reports that a lab's model has produced a complete proof would push it toward Yes.
- Announcements of only partial progress, or lab statements crediting outside mathematicians, would not move resolution toward Yes.
- Public errors found in a claimed proof, or a lab withdrawing a claim, would push the market toward No, though the rules do not say whether a later retraction undoes an earlier qualifying announcement.
- Time passing without any claim makes Yes less likely as the deadline nears.
Put a bot on this market
A PROWL bot reads the news around the clock, judges each post against these exact rules, and trades inside the limits you set. It starts in paper mode.
Questions people ask
Does the AI need to be verified as correct for the market to resolve Yes?
The rules require an announcement by an eligible lab that the problem has been solved. They do not require confirmation from the Clay Mathematics Institute or any other body. They also do not say what happens if the claim is later disputed.
When does the Polymarket AI math problems market end?
The deadline is December 31, 2026, 11:59 PM ET. That is 2027-01-01 04:59 UTC. This matches the end date in the market data.
Does it count if mathematicians solve a problem using an AI model?
Not if the announcement only credits those mathematicians. The announcement must present the solution as the work of the lab, its researchers or its models, alone or jointly with outside researchers.
Which companies count as eligible AI labs?
The rules list many labs, including Anthropic, Google, OpenAI, Meta, Microsoft, DeepSeek, Nvidia, Harmonic and Axiom Math. Subsidiaries and research divisions count too. So does any company with a model on the MMLU-Pro leaderboard when it makes the announcement.
Does partial progress on a problem like the Riemann Hypothesis count?
No. Announcements of partial results or progress toward a solution do not qualify. The announcement must say the problem has been solved.
Is each problem a separate bet?
Yes. Each listed problem is its own market, and each resolves on whether an eligible lab announces a solution to that specific problem by the deadline.
More Science markets explained
Written from the market's own rules text on Nov 17, 2026. If the rules on Polymarket differ from this page, the rules on Polymarket govern.