NVIDIA launches Nemotron for Science with chemistry dataset to aid research

What happened
On November 16, 2026, NVIDIA launched Nemotron for Science, a platform that combines the Nemotron 3.5 Super VL model with the LitmusBench v0.2 chemistry dataset. The system is designed to help researchers build AI agents for scientific tasks, including drug metabolism prediction and laboratory protocol generation. Research institutions such as Argonne National Laboratory and Jülich Supercomputing Centre are adopting the platform for their scientific projects. The tools and data are available on Hugging Face and can be run on local infrastructure like the DGX Spark system.
Key details
- Nemotron 3.5 Super VL model has 120 billion total parameters and supports text, image, and video inputs.
- LitmusBench v0.2 contains nearly 50,000 chemistry questions and is licensed under CC-BY-SA-4.0 and OpenMDW-1.1.
- The platform supports reinforcement learning with verifiable rewards using experimental data as automated verifiers.
- Terray Therapeutics used Nemotron to recover about 76% of annotated chemical reactions in a benchmark test.
- Lila Sciences reported nearly double the accuracy for quantum dot protocol generation compared to GPT-6 Astra, at significantly lower inference cost.
Why it matters
Researchers and scientific organizations can now use Nemotron for Science to automate and improve tasks such as chemical extraction, protocol generation, and therapeutic design. The platform enables these groups to run AI models on their own infrastructure, reducing reliance on external cloud providers and lowering costs. This gives research teams more control over their data and workflows, and provides new tools for advancing scientific discovery in fields like chemistry and drug development.

