What changed
Deep Principle released Hydrogen 1.0, a 753-billion-parameter scientific model post-trained from GLM 5.3 for chemistry, materials and life sciences. Gated BF16 weights carry an MIT license, with NVFP4 and MXFP4 variants for NVIDIA Blackwell and AMD Instinct hardware.
Why it matters
Researchers gain inspectable weights and a documented eight-GPU SGLang path for a domain model, while the separate quantizations make exact-hardware validation possible rather than assuming one artifact behaves everywhere.
What to do
Download only the variant that matches your hardware, preserve generation_config.json, reproduce a small domain benchmark and require literature, calculation or experimental review before any scientific decision.
Caveat
Training datasets and full recipes are not released. Benchmark and discovery results are developer evaluations, the files require account-information sharing, and no independent scientific validation was found.