OpenAI Just Solved a 90-Year-Old Math Problem in 88 Hours — And It’s Way Bigger Than You Think

 

OpenAI Just Solved a 90-Year-Old Math Problem in 88 Hours — And It’s Way Bigger Than You Think

**OpenAI claims its unreleased AI model has cracked one of the seven "Millennium Prize Problems" — a mathematical puzzle that has stumped the world's brightest minds for nearly a century. It took 10,000 AI agents just 88 hours to do it.**

## The $1 Million Question (Literally)

Let me tell you about a math problem so hard, it comes with a $1 million prize from the Clay Mathematics Institute.

It's called the **Navier-Stokes existence and smoothness problem**, and it's been taunting mathematicians since the equations were first written down in the 19th century . The problem basically asks: do these equations that describe how fluids move — water, air, blood — always work perfectly, or can they break down under certain conditions?

For 90 years, nobody could answer it .

Then OpenAI came along.

On September 8, 2026, the company announced that one of its internal AI models had cracked it. The model is reportedly "significantly more capable" than even its latest GPT-6 Astra system . And it took just 88 hours to produce a solution .

## Here's How They Actually Did It

This wasn't just a single AI doing some quick math. Here's what actually happened behind the scenes :

- **September 1, 2026:** OpenAI starts training a new, more powerful model. They hear rumors that two mathematicians are getting close to solving one of the Millennium Problems.

- **September 2:** They decide to unleash the AI on all six remaining unsolved Millennium Problems.

- **September 5:** After 88 hours of work, the AI agents crack the Navier-Stokes problem.

- **September 8:** OpenAI makes the announcement.

The scale is mind-blowing. OpenAI deployed roughly **10,000 AI agents** working in parallel on this single problem . They exchanged nearly 3 million messages and used up 130 billion output tokens — basically, the AI equivalent of a massive brainstorming session that never sleeps .

The computing cost alone was "emphatically in the millions of dollars," according to OpenAI's chief research officer Mark Chen. One estimate put it at about **$15 million** if a customer had run it commercially . OpenAI doesn't plan to claim the $1 million prize .

## What This Actually Means

Here's the fascinating part. The AI proof showed that under certain conditions, the equations that describe fluid motion can actually **break down** — they can produce infinite speed in a finite amount of time. That's a "blow-up" in mathematical terms .

OpenAI researcher Ven Chandrasekaran explained it in plain English: "Our proof does show that there exist fluids which start out perfectly normal, and under the Navier-Stokes equations, actually achieve infinite speed in a finite amount of time" .

Now, because that's physically impossible for a real fluid, it suggests that under certain circumstances, these equations might not always reflect reality accurately. This isn't just math for math's sake — it could have implications for everything from aircraft design to weather forecasting .

## The Drama Behind the Scenes

Here's where it gets messy.

Just hours before OpenAI's announcement, two mathematicians — **Tristan Buckmaster** from New York University and **Levent Alpöge** from Anthropic — released their own AI-assisted work on a related problem . They had apparently been working on it for months.

Buckmaster then claimed that OpenAI only started working on the Navier-Stokes problem *after* information about his research spread. He said the company pursued the same unusual approach that he and Alpöge had been developing .

OpenAI denied accessing their work. "We did not use their prompts or proofs to prompt our models or direct our agents," said researcher Sebastien Bubeck .

But there's a twist. OpenAI later acknowledged: "While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models" .

In other words: the researchers used OpenAI's tools for their work, and that data might have been used to train OpenAI's models . It's the kind of messy intellectual property question that's going to come up a lot more as AI gets smarter.

## What the Math World Thinks

The Clay Mathematics Institute, which administers the Millennium Prizes, hasn't accepted the solution yet. The process is deliberately slow — a proposed solution must be published in a peer-reviewed journal and then survive two years of scrutiny by the mathematical community before the institute even convenes a committee .

Professor Martin Bridson, president of the Clay Institute, told reporters: "This is certainly an exciting day, as we contemplate the announcement of major advances in the human understanding of mathematics" .

Terence Tao, widely considered one of the finest mathematicians alive, called the related work by Buckmaster and Alpöge a "remarkable achievement" on social media . But he's also publicly warned that AI could weaken human understanding of math by solving problems without teaching the process of discovery .

"When mathematicians prove something, they often learned a lot in the process," Tao said. "AI solves problems without really getting any value out of them" .

## Why This Actually Matters

This is the most dramatic sign yet that AI is fundamentally transforming higher mathematics . And it's happening *fast*.

**In May 2026**, an OpenAI model cracked a decades-old conjecture by mathematician Paul Erdős . **A few months later**, Anthropic's Claude model disproved the Jacobian conjecture, which had stood for nearly a century . **Last week alone**, an AI model formalized Fermat's Last Theorem in just 11 days .

OpenAI's latest breakthrough is different. It's tackling a problem that has been a "lighthouse" for mathematicians for generations — a problem that comes with a $1 million prize and has resisted every attempt to solve it for 90 years .

## The Bottom Line: Math Will Never Be the Same

We're witnessing a shift in how knowledge is discovered. For centuries, mathematics has been a human endeavor — a pursuit of pure reason, creativity, and insight . Now, machines are doing it faster, cheaper, and potentially better.

OpenAI's 88-hour solution to a 90-year-old problem is a landmark moment. But it's also a warning: the pace of AI progress in mathematics is accelerating faster than anyone expected.

As OpenAI researcher Sebastien Bubeck put it: "This is the spectacular culmination of the arc we have seen over the last 12 months" . And given how quickly the field is moving, the next 12 months might be even more spectacular.

---

## Frequently Asked Questions (FAQs)

**1. What exactly did OpenAI solve?**

OpenAI claims to have solved the Navier-Stokes existence and smoothness problem — one of seven Millennium Prize Problems established by the Clay Mathematics Institute in 2000. It asks whether the equations that describe fluid motion can break down under certain conditions. OpenAI's proof says yes, they can .

**2. How long did it take?**

The AI model produced the solution in roughly 88 hours after the project started .

**3. How did they do it?**

OpenAI deployed up to 10,000 AI agents working in parallel, using a next-generation model significantly more capable than GPT-6 Astra. The agents exchanged millions of messages and the computing cost was in the millions of dollars .

**4. Is this the first Millennium Problem solved by AI?**

Yes. Only one of the seven Millennium Problems has been solved before — the Poincaré conjecture, solved by mathematician Grigori Perelman in 2003. The Navier-Stokes problem had remained unsolved for about 90 years .

**5. Will OpenAI get the $1 million prize?**

No. OpenAI has said it does not intend to claim the prize. The company says its goal is to demonstrate AI progress, not to collect the reward .

**6. Has the math community accepted the proof?**

Not yet. The Clay Mathematics Institute has not certified the solution. Under the rules, a proposed solution must be published in a peer-reviewed journal and survive two years of scrutiny before the institute will consider it .

**7. Is there a controversy?**

Yes. Hours before OpenAI's announcement, two mathematicians — Tristan Buckmaster (NYU) and Levent Alpöge (Anthropic) — released similar work on a related problem. Buckmaster has raised concerns about whether OpenAI started working on the problem after learning about their research .

---

## Disclaimer

*This article is for informational and educational purposes only and does not constitute professional, legal, or financial advice. The claims described are based on OpenAI's announcement and have not been independently verified by the Clay Mathematics Institute or the broader mathematical community. The Navier-Stokes problem remains listed as unsolved on the Clay Institute's official website until the proposed solution undergoes the standard review process. The author is not affiliated with OpenAI, Anthropic, or any other entity mentioned in this article.*