Editorial illustration for OpenAI Claims Its Model Solved Over 100 Math Problems in a Month
OpenAI Model Solves 100+ Math Problems in Month
OpenAI says an internal model trained for roughly a month cracked more than 100 long-standing math problems, including a second Millennium Prize Problem on top of Navier-Stokes. Training started August 28, according to the company, and the pace apparently caught OpenAI's own mathematicians off guard. Internal discussions have reportedly moved on to a different question: how much warning academic researchers need before results like this land on them.
The claim arrived alongside news that OpenAI is setting up an independent advisory group stocked with mathematicians, including Fields Medalist Timothy Gowers. The timing isn't incidental. Mathematicians have been raising concerns that AI-generated proofs could erode conceptual understanding of the field, even when the answers check out. OpenAI's response is to let outside researchers weigh in on how findings get shared with the field and the public.
But the group's mandate has a limit. OpenAI has kept the speed of its own research off the table, meaning the advisory board can shape disclosure but not the rate at which new results show up. What exactly the model solved, and how, remains unclear.
According to OpenAI, training only kicked off on August 28, putting the entire run at about a month. Even the company's own mathematicians were "surprised" by how fast things moved, and internal conversations have shifted to how to give the academic community enough warning to prepare. Among the solved problems is reportedly a second Millennium Prize Problem, the Hodge conjecture.
Why this matters
We'd take the "100 problems in a month" claim with the same skepticism OpenAI itself invited by admitting it only chased the Millennium Prize Problem after catching wind that Anthropic-linked researchers were closing in. That timeline detail matters more than the headline number. For founders and researchers, the lesson isn't that math is "solved," it's that competitive pressure between labs is now shaping which problems get attention and how fast results get announced.
The Institute for Advanced Study advisory group, with Timothy Gowers on board, is a real check on that dynamic, but it's also a hedge: bringing in outside mathematicians to vet claims only makes sense if you're worried the claims won't hold up on their own. Developers building on these models should watch whether solutions come with verifiable proofs or just plausible-looking output. And anyone pitching AI-for-research tools to investors should note that "solved" is doing a lot of work in that sentence, work that a credentialed skeptic like Gowers is being paid to scrutinize.
Related Reading
Browse all OpenAI news →
Research & Benchmarks Microsoft AI Chief: "If It Isn't Safe We Shouldn't Build It
Research & Benchmarks Google's FACTS benchmark shows 70% factuality ceiling across four tests
Research & Benchmarks Nvidia's DLSS 4.5 beta adds 6x Multi Frame Generation for RTX 50 GPUs
Related Topic OpenAI researcher quits, citing distrust over ad‑driven engagement metrics
LLMs & Generative AI
Related Topic OpenAI launches GPT-Image 1.5 with precise editing for enterprise visuals
LLMs & Generative AI
From Archives AI Malware Found Using Autonomous Hive Mind
From Archives NVIDIA's SoL-Pi Cuts AI Coding Agent Token Traffic by Up to 44.7%
From Archives OpenAI Reveals Three-Part Process for Handling AI Model Reports
Open Source
From Archives ChatGPT Crafts Childbirth Course, Saves Hours of Research
Open Source
Common Questions Answered
How long did it take OpenAI's internal model to solve over 100 math problems?
According to OpenAI, the model was trained for approximately one month, with training starting on August 28. The company's own mathematicians were reportedly surprised by the speed at which the model achieved these results, indicating the pace exceeded internal expectations.
Which Millennium Prize Problem did OpenAI's model reportedly solve?
OpenAI's internal model solved the Hodge conjecture, which is identified as a second Millennium Prize Problem. This achievement was particularly notable as it represents progress on one of mathematics' most challenging unsolved problems.
Why is OpenAI concerned about giving academic researchers advance warning of these results?
OpenAI's internal discussions have shifted to determining how much warning the academic community needs before results like this are publicly announced. The company recognizes that sudden announcements of major mathematical breakthroughs could catch researchers off guard and disrupt ongoing research efforts.
What does the article suggest about competitive pressure between AI labs affecting research priorities?
The article indicates that OpenAI only pursued solving the Millennium Prize Problem after learning that Anthropic-linked researchers were closing in on similar results. This timeline detail demonstrates that competitive pressure between labs is now shaping which problems receive attention and how quickly results get announced publicly.
Further Reading
- OpenAI says unreleased AI model solved 100 math problems in less than 1 month - India Today
- OpenAI's model solves 100+ maths problems, company forms advisory group - Business Standard
- OpenAI says 10,000 AI agents cracked one of math's hardest problems - Phys.org
- On the Navier-Stokes Millennium Prize Problem - OpenAI
- AI Has Solved One of Math’s $1 Million Millennium Prize Problems - Quanta Magazine
