# Altman Reveals OpenAI's New Model: Internal Test Cracked Navier-Stokes Equations, Cost Him Two Sleepless Nights

- Source: 新智元 · 微信公众号
- Published: 2026-10-29T23:42:06.000Z
- RoboHot: https://robohot.news/en/items/fopf3qaefv7wk2tc2x7sml6hi
- Original: https://mp.weixin.qq.com/s?__biz=MzI3MTA0MTk1MA==&mid=2652739064&idx=2&sn=59ca36b46f2cb5778fa8e8a915ae3e3e&chksm=f07e7aa3dc1117398899d287563a99bfdbf52e785e9fe97d314c9a401a7d326c8281335ef1f8&scene=126&sessionid=1793370622#rd

## Summary

On a podcast, Altman revealed that an unreleased OpenAI base model, during an internal "hold-my-beer" test weekend, successively cracked multiple interdisciplinary problems, including the Navier-Stokes equations, and that he lost sleep for two consecutive nights over it. He said such breakthroughs rely on tens-of-millions-of-dollars-scale test-time compute, and predicted AI will move toward RSI, agent team collaboration, and robots building their own data centers.