Deep Principle releases open-weight science model Hydrogen-1.0, beating GPT-6-Astra on FrontierScience
dair_ai · x · 2026-11-03
Deep Principle introduced Hydrogen-1.0, its first open-weight model for scientific research, post-trained from GLM-5.3 on internal scientific data, lab records, and discovery trajectories.
- Scores 38.3% on FrontierScience-Research (multi-step physics/chemistry/biology problems) vs GPT-6-Astra's 33.3% — at lower average output cost per problem; GLM-5.3 base sits near 24%, Kimi K3 near 30%.
- Hits 93.0% on GPQA-Diamonds.
- The team argues science should be open and collaborative, and that open weights are the path to open scientific superintelligence.
- dair-ai: specialized models beat general ones, and this matters most for science.
Related event: Deep Principle Open-Sources Science Model Hydrogen-1.0, Beating GPT-6-Astra(4 posts)→
More from Models
- "Sonnet 5.5 used to be smarter": users sense capability regression — nlarusstone · 2026-11-03
- Apple Intelligence lands in China in December, powered by Alibaba's Qwen instead of ChatGPT — winer666 · 2026-11-03
- Only recent models seem to know "weight surgery" — are labs building Bayesian weight priors? — Tkaraletsos · 2026-11-03
- Victor Taelin floats labs replacing backprop with models editing their own weights at scale — arthurcolle · 2026-11-03
- Custom fine-tunes now beat frontier models at ~1/20th the cost, says investor — bindureddy · 2026-11-03
- OpenAI updates Model Spec: intent over literal instruction-following — sven_ai · 2026-11-03