OpenAI claims its AI solved a 90-year-old math problem in just 88 hours.

OpenAI mobilized approximately 10,000 AI agents to solve the Navier–Stokes problem within 88 hours, but the results have not yet been independently verified and are not considered a complete solution to the Millennium Prize problem.

On September 9, 2026, at 4:10 AM, the BBC reported that OpenAI revealed it had made significant progress in solving an advanced mathematical problem that had remained unproven for over 90 years. Using a new artificial intelligence (AI) model and approximately 10,000 AI agents working together for just 88 hours, they achieved this feat.

ChatGPT's developer said on Tuesday that it has deployed AI agents capable of automating certain tasks to solve problems related to the Navier–Stokes equations, which describe fluid motion and are among the most famous and difficult mathematical problems.

Some key aspects of this problem had lacked complete mathematical proof for approximately 90 years, with OpenAI calling the recent results a landmark and viewing them as evidence of the rapidly evolving capabilities of AI tools.

However, OpenAI's work has not yet been independently reviewed or formally accepted by the Clay Mathematics Institute, the US organization that oversees the Millennium Prize Problems and awards prizes to those who successfully solve these important mathematical problems.

OpenAI stated that the company began training a new AI model in late August and quickly found that the model possessed a high level of mathematical capability.

Currently, this model is still an internal tool used exclusively by OpenAI. OpenAI states that it is "significantly superior" to the latest AI models the company has publicly released. Therefore, researchers decided to test the model on well-known advanced mathematical problems.

OpenAI stated that on September 1st, the company learned of news that two of the Millennium Prize problems might have already been solved. Therefore, they decided to use thousands of AI agents running on a new in-house model to try and solve the remaining problems.

By September 5th, or approximately 88 hours after around 10,000 AI agents began operating, OpenAI stated that it had found a solution to the Navier–Stokes existence and smoothness problem, a key issue related to turbulent flow phenomena, or turbulence, which to date still have many aspects that scientists have not been able to fully explain.

Although it only takes 88 hours, the process requires a tremendous amount of computing resources.

OpenAI revealed that AI agents exchanged nearly 3 million messages and generated a total output of approximately 130,000 billion tokens specifically for solving Navier–Stokes problems.

Based on the pricing OpenAI currently charges for the output of its most advanced models, such an effort could cost around $10 million, or approximately £7.3 million.

However, the work that OpenAI says can be proven only covers 2 out of the 4 statements that need to be proven according to the Millennium Prize criteria, which has a prize of 1 million dollars for whoever can solve the problem.

OpenAI emphasized that the purpose of this release is to report significant progress on its AI models, not to claim the Millennium Prize.

OpenAI's announcement sparked controversy after Tristan Bugmaster, a mathematics professor at New York University (NYU), revealed that he and Levent Alperge, a mathematician working on Anthropic, OpenAI's competitor, were also trying to find a solution to the same problem.

Both used OpenAI's Codex as one of the tools for their work, but Bugmaster noted that on September 3rd, he learned that "information about the progress of our work had been sent to OpenAI."

Bugmaster claims that OpenAI did not begin working with the Navier–Stokes equations until after information about his and Alperge's work reached the company, revealing excerpts from emails exchanged with OpenAI to address questions about the timeline and methodology of the company's operations.

OpenAI congratulated BugMaster and Alperge on their work, which was completed around the same time, calling it remarkable. The company stated that it had not seen either of their works through any channels before they were publicly disclosed and confirmed that it did not access user data while working on the Navier–Stokes problem.

However, OpenAI states that, although unlikely, it cannot rule out the possibility that decrypted data obtained from the use of both companies' products may contribute to improving its model.

OpenAI emphasizes that its proof differs significantly from the work of Bugmaster and Alperge, and even the results that each side proves are different.

Therefore, while OpenAI states that the latest results represent a significant advance in AI for solving advanced mathematical problems, these results still need to be validated by independent mathematicians and are not yet considered a complete solution to the Navier–Stokes Millennium Prize.

Reference: www.bbc.com

Read news related to All situations around the world can be found here.