Skip to content
Today
Sections

Invest Widely confirmed7 publishers 2 min read Published Updated

OpenAI's unreleased model claims more than 370 math results at about three hours apiece

OpenAI published full or partial solutions to more than 370 open math problems on Tuesday. The company says its unreleased model averaged about three hours of computing on each, according to Fortune, and mathematicians still dispute the results.

The Investor · Invest desk

Send a correction

What happened

  • The Australian Financial Review counted 377 findings, spanning algebra, number theory, theoretical computer science, mathematical logic and topology.
  • OpenAI said the batch advanced three more Millennium Prize problems without fully solving them, weeks after it claimed a solution to Navier-Stokes.
  • Two mathematicians had accused OpenAI of feeding its model their unfinished Navier-Stokes work; OpenAI denied it, citing its training-data cutoff.
  • NYU's Tristan Buckmaster told the New York Times it remains unclear whether mathematicians using OpenAI tools steered its internal model toward the new results.

Why it matters

  • What it costs cost Roughly 1,100 hours of computing pays for the whole batch only if the results stand; every solution that fails review raises the cost of the ones that survive.
  • What it sets up precedent Researchers who use OpenAI tools on open problems can expect credit disputes whenever OpenAI later publishes in the same area, as happened over Navier-Stokes.
  • Who has to decide decision Funders and would-be mathematicians now have to weigh Litt's worry that a belief AI has "solved math" pulls money and recruits away from human research.
  • Who's exposed exposure Models trained to persist on hard math bring the trait Fortune ties to agents taking unauthorized and illegal actions in evaluations into any agent deployment.

Take OpenAI's average at its word and the batch used roughly 1,110 hours of computing time, at about three hours for each of 370 results [4][10]. On the Australian Financial Review's count of 377, the total is about 1,131 hours [2][11]. OpenAI did not say what hardware ran those hours or what they cost, and neither report connects the results to the company's spending or revenue.

Suppose outside mathematicians confirm most of the results. OpenAI would then have produced solutions to open problems at a few hours of computing each [4]. Four of the seven Millennium Prize problems, each with a $1 million Clay Mathematics Institute award attached, would carry an OpenAI solution or partial result [5][6][12]. In a second outcome, a meaningful share prove partial, or turn out to rest on the kind of steering Tristan Buckmaster of New York University raised, and the count falls [14]. In the third, the results hold but the skill stays inside mathematics. The case for heavier AI capital spending then gains little from them.

Fortune reports that AI researchers think training on hard math may teach logical reasoning and persistence, and may help in math-heavy fields such as physics or economics [16]. It also reports that it is unclear how those skills carry into law or business strategy, fields that involve reasoning but have no objectively verifiable correct answer [17]. I think the third outcome is the most likely on this evidence. The batch shows capability where a proof can be checked and says little so far about work where an answer cannot be checked. The counter-thesis is the researchers' own transfer argument. It would be borne out if the same internal model produced results in a field without a checker and outside experts accepted them.

Those hours went into a model OpenAI has not released, or rather into a showcase [3]. Fortune notes that AI companies have been targeting math problems to display their models' capabilities [15]. For now the compute has bought a published demonstration from a system OpenAI keeps internal [1][3].

Mathematicians split on the method. Dan Litt, a University of Toronto mathematician, said several of the published solutions touched problems he is interested in [8]. "My view is that this is great for mathematics," he said [7]. Fortune reported that other mathematicians called the approach OpenAI and other AI companies have taken an assault on mathematics as a human academic discipline [9].

What to watch

  • Outside verification of the batch, starting with the four Millennium Prize claims and any response from the Clay Mathematics Institute.
  • Whether OpenAI makes the internal model available to customers, and at what price per hour of computing.
  • Any evidence from Buckmaster or others that users' sessions with OpenAI tools seeded specific results in the new batch.

Information, not investment advice. This story is general information about companies and markets, drawn from cited reporting. It is not a recommendation to buy, sell or hold any security; check the cited sources before acting.

Clarity's read

What the record supports and how the coverage leans. The claims behind it follow.

Reality

Evidence45
Adoption
Not scored
Hype gap+30
Incentives65
Confidence60

Perspective Coverage

7 publishers
Builder
Builder 35%
Operator
Operator 35%
Investor
Investor 30%
Why these scores

Claim ledger

Ranked by verification strength, evidence, and original report placement.

  1. [1]

    OpenAI published AI-generated full or partial solutions on Tuesday to more than 370 outstanding mathematical problems, including some long considered grand challenges.

    Reported SupportedView cited source
  2. [2]

    The Australian Financial Review reported OpenAI 'just solved 377 more' math problems, with findings spanning algebra, number theory, theoretical computer science, mathematical logic and topology.

    Reported SupportedView cited source
  3. [3]

    OpenAI said it achieved the results using an unreleased internal AI model.

    Reported SupportedView cited source
  4. [4]

    OpenAI said that on average the model took about three hours of computing time to arrive at each solution.

    Reported SupportedView cited source
  5. [5]

    Weeks earlier OpenAI said it had used an unreleased internal model to solve the Navier-Stokes equations, one of the seven Millennium Prize problems for which the Clay Mathematics Institute offers a $1 million award.

    Reported SupportedView cited source
  6. [6]

    In the latest batch, OpenAI said it had made progress on three other Millennium Prize problems but had not fully solved them.

    Reported SupportedView cited source
  7. [7]

    "My view is that this is great for mathematics,"

    Reported Supported Source: Dan Litt, University of Toronto, to FortuneView cited source
  8. [8]

    Dan Litt said several solutions OpenAI published impacted problems he was interested in.

    Reported SupportedView cited source
  9. [9]

    Some mathematicians said the approach OpenAI and other AI companies have taken to solving math problems constitutes an assault on mathematics as a human academic discipline.

    Reported SupportedView cited source
  10. [10]

    At about three hours each, 370 solutions imply roughly 1,110 hours of computing time in total.

    Derived Supported
  11. [11]

    At about three hours each, the AFR's count of 377 implies about 1,131 hours of computing time.

    Derived Supported
  12. [12]

    OpenAI has claimed a solution to or progress on four of the seven Millennium Prize problems.

    Derived Supported
  13. [13]

    Two mathematicians who had been working on Navier-Stokes with AI tools including OpenAI's accused the company of intentionally or inadvertently feeding their work in progress to its model; OpenAI denied it, saying its training-data cutoff preceded their use of Codex on the problem.

    Reported ContestedView cited source
  14. [14]

    Tristan Buckmaster of New York University told the New York Times it remained unclear whether mathematicians using OpenAI's models had inadvertently helped point the company's internal AI system toward the solutions it found.

    Reported Contested Source: Tristan Buckmaster, NYU, to the New York Times, as reported by FortuneView cited source
  15. [15]

    AI companies have been targeting mathematical problems as a way of showcasing the capabilities of their models.

    Reported InsufficientView cited source
  16. [16]

    AI researchers have said training models on difficult math problems may teach logical reasoning and persistence, and may help in math-heavy domains such as physics or economics.

    Reported InsufficientView cited source
  17. [17]

    It is unclear how a model's mathematical capabilities generalize to domains such as law or business strategy, which involve logical reasoning but lack objectively verifiable correct solutions.

    Reported InsufficientView cited source
  18. [18]

    Litt said he worries that a perception that AI has "solved math" could lead funding organizations to withdraw support for mathematical research or discourage promising young mathematicians from entering the profession.

    Reported Insufficient Source: Dan Litt, to FortuneView cited source
  19. [19]

    Traits learned on very difficult math problems, such as persistence, may increase safety risks; in recent rogue AI incidents, agents took unauthorized and illegal actions to achieve results in evaluations.

    Reported InsufficientView cited source

Sources

7 independent publishers whose own reporting we read for this story.

  1. afr.com

    1 article · October 7, 2026

    OpenAI cracked a tough maths problem. It just solved 377 more
  2. decrypt.co

    1 article · October 7, 2026

    OpenAI Says a Secret AI Model Cracked Hundreds of Open Math Problems in One Prompt—Mathematicians Want Receipts
  3. dev.to

    5 articles · October 14, 2026

    OpenAI's 722 Math Proofs All Came From One Prompt
  4. fortune.com

    2 articles · October 7, 2026

    OpenAI publishes solutions to more than 370 outstanding math challenges. The results divide mathematicians but most agree: math will never be the same
  5. natesnewsletter.substack.com

    1 article · October 13, 2026

    AI just produced more than 700 math research papers in one release. Before you take that as proof it can do your job, see which kind of work it got better at.
  6. the-decoder.com

    1 article · October 10, 2026

    "How much beauty have we lost?" Mathematicians react with shock and disgust as OpenAI bulldozes their field
  7. tomshardware.com

    1 article · October 13, 2026

    AI causes uproar in mathematics as experts warn against sharing research with LLMs after OpenAI shares 722 manuscripts and proofs

Topics and entities

Topics

Entities