Which open math problems has AI actually solved so far?
AI has actually solved some open math problems, but the verified list is still small and depends a lot on what you count as “solved.”
The clearest cases
Overview
What has been solved?
Confirmed
9 Erdős open problems
44 OEIS conjectures
a disproof of a discrete geometry conjecture about unit distancesofficial announcements of AI-assisted or AI-generated progress
a 40-year-old open optimization probleminvolving Nesterov accelerated gradient methods, solved with AI assistance
Claimed, not settled
Navier–Stokes existence and smoothness problemannounced claims awaiting independent mathematical verification and publication
Not confirmed
Riemann Hypothesis
P vs NP
Drawn by consens.io from this answer of Oct 15, 2026. Every date, number and name is taken from it.
1) A set of open problems in combinatorics/discrete math
The strongest documented successes are AI-generated proofs that were machine-checked and then publicly released for review. The most concrete recent report says an AI-driven formal proof system solved:
9 Erdős open problems
44 OEIS conjectures
These are real open problems, not contest problems. The evidence is strongest when the proof is formalized in a proof assistant such as Lean, because that gives a mechanically checked proof rather than just a plausible-looking writeup.
2) Specific discrete-geometry and optimization results
There are also official announcements of AI-assisted or AI-generated progress on individual long-standing problems, including:
a disproof of a discrete geometry conjecture about unit distances
a 40-year-old open optimization problem involving Nesterov accelerated gradient methods, solved with AI assistance
several new mathematical constructions and bounds in combinatorics and coding theory
These are genuine mathematical advances, though not all of them are “full solution of a famous big conjecture.”
3) New constructions and improved bounds
AI has also solved open problems in the sense of finding new best-known constructions, algorithms, or bounds, for example:
cap-set constructions
new matrix-multiplication / algorithm-design results
new bounds in coding theory and combinatorics
These count as mathematical progress, but not always as a complete resolution of the underlying open problem.
What is still unconfirmed or provisional
Some very dramatic claims have been announced recently, including a purported AI solution to the Navier–Stokes existence and smoothness problem and a broad batch of results across several famous conjectures. As of now, those should be treated as announced claims awaiting independent mathematical verification and publication, not as universally accepted solved problems.
Bottom line
If you mean “Which open problems has AI definitely solved in a way that’s been checked?”, the safest answer is:
some specific Erdős problems
some OEIS conjectures
some discrete geometry / optimization problems
some construction-and-bound problems in combinatorics, coding theory, and algorithms
If you mean famous grand-challenge problems like the Riemann Hypothesis, P vs NP, or Navier–Stokes, then none are yet broadly accepted as solved by AI.
Only when a check moves the answer. No account needed.
Double opt-in · unsubscribe anytime · your address is only used for these updates.
History
Checked 28 times since Jul 25
How strongly the models agree, per check. Amber points mark checks that changed the answer.
The answer changed
The new answer adds the KLS conjecture as a major AI-assisted problem reportedly proved in 2026, supported by a newly surfaced Microsoft Research source. It also names additional specific results from OpenAI’s mathematics release.
The new answer adds documented AI-generated/formally checked resolutions of open Erdős problems, including a public writeup and broader proof-search results. This strengthens the evidence beyond the standing answer’s framing of outright open-problem solutions as mostly recent claims awaiting broad verification.
The consensus shifted from claiming AI has actively resolved major open problems (like Erdős problems, Navier-Stokes, and Fermat's Last Theorem) to a much more cautious stance that AI has only achieved silver-medal Olympiad levels and improved bounds/constructions, while claims of solving famous open problems lack sufficient independent validation.
The new consensus substantially strengthens the conclusion, asserting that AI has settled major open problems, including a Navier–Stokes result, and resolved numerous Erdős and other conjectures. The old consensus treated such claims as tentative, unverified, and not community-accepted, while affirming only material advances and possible machine-checkable results.
The nine Erdős results have been published in Science after peer review.
The nine results underwent peer review and were published in Science.
Gemini
“these solutions went through a standard, rigorous peer-review process and were published in the journal *Science*”
The results are reported in a preprint and are not yet broadly accepted in the literature.
OpenAI
“This is reported in a 2026 research preprint, so it is strong computational evidence but not yet equivalent to broad human acceptance in the literature.”
How to verify: Check the publication status and peer-review history of the report describing the nine Erdős results.
Contradiction · critical
The Navier–Stokes problem has been definitively solved by AI.
OpenAI’s system produced a solution resolving the Navier–Stokes Millennium Problem.
Grok
“An internal OpenAI system produced a solution to the Navier–Stokes existence and smoothness Millennium Prize Problem”
The announced solution remains provisional pending independent mathematical verification.
OpenAI
“OpenAI has announced a claimed AI-generated solution to Navier–Stokes, but it should not yet be reported as an uncontroversially established solution.”
How to verify: Check whether specialists have independently validated the proof and confirmed that its formalized statement matches the Millennium Problem.
