ai lab
Also indexed as periodic · periodiclabs
Periodic Labs
Periodic Labs develops scientific AI alongside autonomous physical laboratories, including the Neon model announced in September 2026.
Research dossier · Reviewed 2026-09-21
Research and release status
Periodic Neon was announced on September 15, 2026 for work in the company's physical labs. It is not yet indexed in this Atlas snapshot.
Learning from physical experiments
Periodic combines model development with autonomous laboratories in the physical sciences. The program uses experimental results, including failures that may never enter scientific publications, as training evidence. Materials discovery is a central application. [1]
Neon changes the release picture
The September 15 announcements identify Periodic Neon as a one-trillion-parameter model post-trained to analyze experimental results and deployed in the company's labs. Related posts cover X-ray diffraction analysis and infrastructure. This is an announced scientific model, not a reason to assume public weights or unrestricted API access. [2]
What establishes scientific value
The important evidence is whether predictions survive physical testing and shorten useful experimental cycles. The company's description of its own results is a starting point; independent replication and downstream discoveries would strengthen the assessment. [1][2]
Sources
- [1] Research program and team ↗Periodic Labs · accessed 2026-09-21
- [2] Neon and research announcements ↗Periodic Labs · accessed 2026-09-21