Ex-OpenAI's Periodic trains 1T-param science model Neon on 1,300 H200s to beat GPT-6 Astra
LiamFedus · x · 2026-09-16
Periodic Labs has unveiled Neon, a 1T-parameter model post-trained to analyze experimental results and deployed in its physical materials labs in Menlo Park, targeting superconductors, magnets and semiconductor materials.
Key facts:
- Founder Liam Fedus (ex-OpenAI) says the company built high-throughput labs creating a loop: experiments generate fresh data, models learn from it and guide what to try next.
- Using only 1,300 H200 GPUs plus months of proprietary experimental data, they mid-trained and RL'd an open-source model to surpass GPT-6 Astra on their own analysis benchmark.
- Neon is framed as an early glimpse of physical labs owning AI that beats closed models and keeps improving from their data.
- Infra built with @radixark, SGLang, NVIDIA and Ray, spanning labs, simulation tools, training/inference/RL sandbox — a full atoms-and-bits stack.
More from Companies & People
- Alpha School investigation: 17 campuses on probation, Bill Ackman as shadow lead, Holmes involved — benjaminjriley · 2026-11-02
- ELLIS names 160 new ML Fellows and Scholars for 2025–2026 — JJitsev · 2026-11-02
- Andrew Ng releases free 3-hour AI agents masterclass: it's about systems, not prompts — hey_abusiddik · 2026-11-02
- Why OpenAI must copy Google's long game to keep up with Anthropic — haider1 · 2026-11-02
- US firms out-extracted Europe on IT via management — same may hold for AI — carlbfrey · 2026-11-02
- Sam Altman: AI query energy cost fell 1,000x-100,000x in 3-4 years — rohanpaul_ai · 2026-11-02