{"slug": "openai-claims-ai-solved-the-90-year-old-navier-stokes-problem", "title": "OpenAI Claims AI Solved the 90-Year-Old Navier-Stokes Problem", "summary": "OpenAI claimed on Tuesday that its internal AI system produced a proof resolving the Navier-Stokes existence and smoothness problem, one of the seven Millennium Prize Problems identified by the Clay Mathematics Institute in 2000, each carrying a $1 million prize. The proof, which shows a singularity can form in fluid motion, was formally verified using Lean, with about 10,000 AI agents generating 2.7 million messages and 130 billion output tokens over 88 hours, plus 17 hours for verification. The mathematical community has not yet validated the result, and OpenAI said it does not intend to claim the $1 million prize.", "body_md": "**September 9, 2026**, (Inside AI) — OpenAI has claimed a breakthrough on one of mathematics' most stubborn open questions. On Tuesday, the company said its internal AI system produced a proof resolving the Navier-Stokes existence and smoothness problem.\n\nThe problem is one of the seven Millennium Prize Problems identified by the **Clay Mathematics Institute** in **2000**. Each carries a **$1 million** prize for a verified solution. The Navier-Stokes question asks whether smooth solutions to the equations governing fluid motion can break down in finite time. It has resisted proof for roughly **90 years**.\n\nOpenAI's claimed proof shows that a singularity can indeed form. The company described a vortex that spirals inward while being stretched. As the central region shrinks, fluid speed increases until the mathematical singularity appears. Crucially, the fluid retains finite energy, meaning the breakdown emerges from the equations' own dynamics, not an external infinite force.\n\nThe proof was formally verified using **Lean**, a computer system that checks mathematical proofs step-by-step. OpenAI said about **10,000 AI agents** worked on the problem simultaneously. They generated **2.7 million messages** and used **130 billion output tokens**. The agents found the result in **88 hours**, with formal verification taking another **17 hours**.\n\nOpenAI also stated the AI model behind the proof is significantly more capable than the newly launched **GPT-6 Astra**. This detail suggests a specialized internal system, not a public product, drove the discovery.\n\n## Caution Before the Prize\n\nThe broader mathematical community has not yet validated the result. OpenAI released the proof for examination but said it does not intend to claim the **$1 million** prize. The company's announcement has already sparked a dispute involving mathematicians **Tristan Buckmaster** of **New York University** and **Levent Alpöge** of **Anthropic**.\n\nBoth had been working on related problems. They raised questions about whether their unpublished work could have influenced OpenAI's research. OpenAI denied accessing their work and said its proof is substantially different.\n\n**\"I have not seen OpenAI's proof. I do not know what their model did, or how. I do not know whether our data was used. I am not accusing anyone of anything,\"** Buckmaster said in a statement.\n\nThe Navier-Stokes equations describe how fluids like water, honey, and air flow. They underpin weather forecasting, aircraft design, and ocean current prediction. The core difficulty is turbulence. Tiny changes in starting conditions can lead to wildly different outcomes. This is why weather cannot be predicted reliably more than a week ahead.\n\nIf the proof withstands scrutiny, it would mark a major milestone. An AI system would have helped resolve a problem that has stumped humans for about a century. But history counsels patience. Millennium Prize claims have collapsed before, and formal verification in Lean does not automatically equal mathematical acceptance.\n\n## What Comes Next\n\nIndependent mathematicians must now dissect the proof. The dispute over research conduct adds another layer of scrutiny. OpenAI's decision to forgo the prize money may reduce financial incentives for challenge, but it does not lower the scientific bar.\n\nThe use of **10,000 AI agents** working in parallel also raises questions about reproducibility. Can other teams recreate the result without the same computational scale? Will the proof be understandable to human mathematicians, or will it exist only as a machine-checked artifact?\n\nFor now, the claim is precisely that: a claim. The Navier-Stokes problem remains officially unsolved until the mathematical community says otherwise.", "url": "https://wpnews.pro/news/openai-claims-ai-solved-the-90-year-old-navier-stokes-problem", "canonical_source": "https://insideai.news/news/machine-learning/navier-stokes-problem-openai/10037/", "published_at": "2026-09-09 11:50:51+00:00", "updated_at": "2026-09-09 17:58:07.670242+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-research", "ai-agents"], "entities": ["OpenAI", "Clay Mathematics Institute", "Lean", "GPT-6 Astra", "Tristan Buckmaster", "New York University", "Levent Alpöge", "Anthropic"], "alternates": {"html": "https://wpnews.pro/news/openai-claims-ai-solved-the-90-year-old-navier-stokes-problem", "markdown": "https://wpnews.pro/news/openai-claims-ai-solved-the-90-year-old-navier-stokes-problem.md", "text": "https://wpnews.pro/news/openai-claims-ai-solved-the-90-year-old-navier-stokes-problem.txt", "jsonld": "https://wpnews.pro/news/openai-claims-ai-solved-the-90-year-old-navier-stokes-problem.jsonld"}}