Eigen RadarAI
Analysis

NVIDIA releases Nemotron for Science with chemistry training data

NVIDIA released Nemotron for Science, combining an open model with chemistry data and instructions for adapting it to specialized research tasks. The package includes nearly 50,000 chemistry reasoning questions and is available on Hugging Face. Research organizations are integrating the model into scientific agents, including an Argonne system for designing and evaluating protein-based therapeutic candidates.

Artificial Intelligence··Morning
Wide view of a medicinal chemistry and drug-discovery laboratory lined with glass fume hoods, reaction flasks and trays of solubility sample vials.

A science model arrives with adaptation tools

NVIDIA introduced Nemotron for Science on November 16, combining the Nemotron 3.5 Super VL open model with scientific evaluations, chemistry data and post-training recipes. The model, dataset and adaptation resources are available on Hugging Face, the platform for sharing AI models and data. Researchers can use specialized data to adapt the model and connect their scientific agents to computational tools.[1], [2]

Chemistry questions support training and evaluation

The LitmusBench v0.2 dataset contains nearly 50,000 chemistry reasoning questions. It draws on molecular properties and experimental drug-discovery data, including solubility and the way the human body processes molecules. NVIDIA provides it for both evaluating and further training models. Adapted agents can interpret scientific information, generate code and call computational tools through the company’s CUDA-X libraries for accelerated scientific calculations.[1]

Research teams integrate the model into laboratory work

The US Department of Energy’s Argonne National Laboratory is integrating Nemotron into StructBioReasoner, its multi-agent system for designing and evaluating protein-based therapeutic candidates. Terray Therapeutics, a drug-discovery company, adapted the model to extract reactants, products and reaction conditions from scientific diagrams. Lila Sciences uses it to prepare executable protocols for automated laboratories.[1]

References

  1. News sourceNVIDIA NewsroomNVIDIA releases Nemotron model and data resources for science agents↩1↩2↩3
  2. News sourceStorageReviewNVIDIA releases Nemotron for Science model and chemistry data↩