Deep Principle Releases Hydrogen-1.0: A 753B Open-Weight Model

Deep Principle has officially announced the release of Hydrogen-1.0, an open-weight scientific intelligence model boasting an immense 753 billion parameters. Engineered to tackle complex challenges in chemistry, materials science, and the life sciences, this new model is designed to supercharge laboratory discoveries and bridge the gap between artificial intelligence and empirical science.

To achieve its specialized capabilities, Hydrogen-1.0 was built by conducting rigorous post-training on Z.ai's GLM-5.3 foundation model. The training regimen utilized specialized domain data, extensive laboratory records, and agent discovery trajectories, enabling the model to process intricate scientific workflows and experimental data far more effectively than general-purpose language models.

Stock photo for illustration only, not from the actual event

For developers and computational scientists, the release provides open access to weights that reportedly achieve a FrontierScience-Research win over GPT-6-Astra. This benchmark performance highlights the model's exceptional capability in navigating rigorous scientific evaluations and solving multi-step research hypotheses that typically require specialized human expertise.

By distributing the model under the MIT License, Deep Principle ensures that academic institutions and commercial entities alike can integrate, modify, and deploy Hydrogen-1.0 into their proprietary research pipelines with minimal legal friction. As open-science initiatives continue to gain momentum, this release marks a pivotal milestone in the evolution of AI-accelerated scientific discovery.

Source: MarkTechPost