Periodic Labs unveils "labs that learn": 1T-param Neon now deployed in its own labs

LiamFedus · x · 2026-09-16

In its blog post "Building Labs that Learn," Periodic Labs describes Neon's three pillars: high-throughput labs, research, and infrastructure. The thesis: experiments from the labs provide the data to train increasingly capable scientific AI, which in turn guides better experiments; when frontier models fall short or cost too much, they train their own. Neon, a 1T-parameter model mid-trained and RL'd on lab data, beats GPT-6 Astra and Claude Fable 5.1 on diffraction analysis and is already deployed in the company's 24/7 labs — a step toward putting autonomous discovery in scientists' hands.

Related event: Periodic Labs beats GPT-6 on materials science with just 1,300 H200s(41 posts)→

Original post →

More from Research

Research channel →