Periodic Labs introduces Periodic Neon, a 1T-parameter lab-trained model that beats GPT-6 Astra at X-ray diffraction analysis
On Sept 15, 2026 Periodic Labs, the materials-science startup of Liam Fedus and Ekin Dogus Cubuk, published its first research. It introduced Periodic Neon, a 1-trillion-parameter model built by mid-training and RL-tuning the open-weight Kimi K2.6 on the company's own laboratory data, using 1,300 H200 GPUs. On the internal FrontierXRD benchmark Neon reached 55.3%, up from the base model's 2.7%. The company says it beats GPT-6 Astra and Claude Fable 5.1 at diffraction analysis, and Neon is deployed in its Menlo Park labs.
Key facts
- 1T parameters; post-trained from open-weight Kimi K2.6 with scientific midtraining + RL on proprietary XRD lab data
- Final training run on 1,300 H200 GPUs
- FrontierXRD (internal eval): Neon 55.3% vs Kimi K2.6 2.7%; company says it surpasses GPT-6 Astra and Claude Fable 5.1 'at a lower cost per analysis' (their scores not given in the post)
- Deployed in Periodic's physical labs to analyse experiments in the search for superconductors and magnets; no discoveries claimed yet
- Three posts the same day: 'Building Labs that Learn', 'Nature Is Our Learning Environment', 'AI Infrastructure at Periodic'
- No announcement of releasing Neon's weights
Science result
- Field
- materials / X-ray diffraction analysis
- Problem
- Automated interpretation of XRD experiments in an autonomous materials lab
- Result
- Lab-data-trained 1T model reaches 55.3% on the FrontierXRD internal benchmark (base model 2.7%), reported above GPT-6 Astra and Claude Fable 5.1
- AI system
- Periodic Neon
- Human role
- Human-led: Periodic researchers trained the model and run the labs
- Verification
- Company-reported on an internal benchmark; not independently replicated
- Status
- pending
What happened
Periodic Labs came out of stealth in 2025 with a $300M seed round. Its first public results show the "lab as RL environment" idea: experimental data from its own autonomous labs is used to specialise a large open-weight model for one scientific task, analysing XRD patterns.
Why it matters
It shows a startup taking a Chinese open-weight frontier model and, with modest compute and proprietary experimental data, beating the best closed models on a narrow scientific task. That is an argument that experimental data, not general capability, is the bottleneck in AI-for-science.
Changelog
- 2026-10-01: created (leads run; covers Import AI 474 "Building Labs that Learn" lead too)
Related events
- Periodic Labs launches with a $300M seed round to build AI scientists with autonomous labs ★★★
- Moonshot AI releases Kimi K3, a 2.8T-parameter open-weights multimodal model ★★★★★
Sources (4)
- officialPeriodic Labs: Building Labs that Learn
- officialPeriodic Labs: Nature Is Our Learning Environment
- officialPeriodic Labs: AI Infrastructure at Periodic
- pressMenlo Times: Periodic Labs introduces Periodic Neon
id: 2026-09-15-periodic-labs-neon · updated 2026-10-01 · open in the interactive timeline