Hydrogen-1.0

Open-weight science model for chemistry, materials, and life sciences, post-trained from GLM-5.3.

Also known as Hydrogen 1.0

What it does

Hydrogen-1.0 is Deep Principle's open-weight scientific model, built by post-training GLM-5.3 on chemistry, materials science, and life sciences, plus laboratory records and agent trajectories from Deep Principle's own labs. The blog says the weights are on Hugging Face and ModelScope, that NeuroSpark offers API access, and that Mira uses the model in Deep Principle's AI Scientist. The Hugging Face model card licenses the BF16 weights under MIT, lists a 1 million token context and a 128K token maximum output, and says independently authored code and documentation are Apache 2.0. The Hub API lists the repository as created on October 31, 2026 and last modified on November 2, 2026. The blog does not print a calendar date. Its assets are versioned v20261102a. In Deep Principle's own evaluations, the blog says Hydrogen-1.0 outperforms GPT-6-Astra on FrontierScience-Research and improves over GLM-5.3 on GPQA-Diamonds, roughly matching Kimi-K3. Those comparisons are the vendor's. The model card says full datasets and recipes are not released with the weights.

Who is Hydrogen-1.0 for?

Researchers who want an open-weight model for chemistry, materials, and life-science questions.

Tags

• open-weight
• science
• chemistry