What does Noam Brown say about OpenAI's 10,000-agent Millennium Prize result?

Noam Brown says the multi-agent setup deserves less than 10% of the credit for OpenAI solving a Millennium Prize problem on Navier-Stokes. In his view the real reason is a very strong general-purpose model that can work over long horizons and think in parallel, and the swarm gets outsized attention because it is new and flashy. He also notes OpenAI has not measured how long a single agent would take or whether 10,000 agents beat 1,000.

• Dwarkesh Patel described the run as 10,000 agents spending 130 billion tokens over 88 hours.
• Brown says experiments at that scale are too expensive to repeat for a proper comparison, so OpenAI studies scaling at 64 to 256 agents instead.

▸In their words

“I wouldn't even attribute 10% of the credit to multi-agent.” Noam Brown · at 6:00 — open in the transcript →

Source: our own transcription of the episode. The full conversation is on Dwarkesh Podcast (2026-09-17). Read the whole episode summary →

▸More from this episode