Deep Principle has released Hydrogen-1.0, its first open-weight model for researchers in chemistry, materials science, and life sciences. The company post-trained Z.ai's GLM-5.3 on domain data, laboratory records, and discovery trajectories from its own labs, with the aim of helping scientists formulate questions, design experiments, and use the resulting evidence to decide what to try next. Weights are published under an MIT license on Hugging Face and ModelScope, with an API through NeuroSpark and access inside Mira, the company's AI Scientist. In the company's evaluations, Hydrogen-1.0 scores 38.3% on FrontierScience-Research, above the 33.3% it reports for GPT-6-Astra, and 93.0% on GPQA-Diamond, improving on GLM-5.3 and roughly matching Kimi-K3. On internal tasks it leads GPT-6-Astra at maximum thinking by 5.13 points in scientific computing and 4.76 points in one-step forward reaction prediction, and trails by 3.33 points on NMR structure elucidation. The launch writeup shows it writing and repairing a polyethylene molecular dynamics workflow in GROMACS, mapping an Ireland-Claisen reaction network, and compiling a 12-vial Suzuki coupling protocol that a virtual rehearsal can correct before a run. The model card lists 753 billion parameters, a 1 million token context window, and a 128K token output limit. BF16 weights were tested with SGLang on eight GPUs, and quantized builds target NVIDIA Blackwell and AMD Instinct GPUs. Founder and CTO Chenru Duan announced the release. With CEO Haojun Jia, he founded Deep Principle to apply AI to materials research, and the company runs a high-throughput laboratory alongside its models. Training combined supervised fine-tuning with reinforcement learning on research outcomes, and the full datasets and recipes are not released with the weights.