Liam Fedus's Lab Builds Neon With Just 1,300 H200s, Beats GPT-6 Astra on Materials Benchmark
LiamFedus · x · 2026-09-16
Former OpenAI VP of research Liam Fedus detailed his new lab's loop: high-throughput materials labs in Menlo Park generate fresh experimental data, models learn from it, then steer what to try next.
- Using only 1,300 H200s plus months of proprietary experimental data, the team mid-trained and RL'd an open-source model to surpass GPT-6 Astra on their analysis benchmark—dubbed Neon
- First targets: hard problems in materials science including superconductors, magnets, and semiconductor materials
- Framed as evidence AI is collapsing the barrier to hardware creation: 'a hardware engineer in their pocket'
- A landmark AI-for-Science case: vertical data + modest compute beating a general frontier model
Related event: Periodic Labs trains Neon on 1,300 H200s to beat GPT-6(54 posts)→
More from AGI Musings
- Periodic Labs beats GPT-6 Astra on X-ray analysis with Neon, trained on just 1,300 H200s — hsu_byron · 2026-09-16
- Most Millennium Problems couldn't even be asked in 1900—AI math lacks that conceptual leap — alexbilz · 2026-09-16
- Nate Silver: We're not ready for superpersistent AI — StefanoGogioso · 2026-09-16
- Multiplayer AI called the biggest design problem of 2027, with interfaces far from settled — manosaie · 2026-09-16
- Unless you plan to tyrannize future generations, posthumans are happening — LesaunH · 2026-09-16
- Altman Tells Benioff: Models Got Good Faster Than Anyone Expected, Power Concentration Is a Real Fear — kimmonismus · 2026-09-16