Navier-Stokes run details: 10k agents, 88 hours, 270M messages, 130B output tokens

daniel_mac8 · x · 2026-09-09

danielmac8 compiled the run details of OpenAI's Navier-Stokes solution: an internal model still in training, 10,000 concurrent agents, 88 hours of runtime plus 17 hours of verification with Astra, 2.7 million messages, and 130 billion output tokens. He argues this presages personal superintelligence 100x beyond GPT-Next on an iPhone-sized device within a decade.

Related event: OpenAI's Navier-Stokes Claim Ignites Community Storm(33 posts)→

Original post →

More from AGI Musings

AGI Musings channel →