NVIDIA Posts Nemotron 3.5 Super VL and 50,000 Question Chemistry Dataset on Hugging Face
Summary
NVIDIA introduces Nemotron for Science, combining its Nemotron 3.5 Super VL open model with LitmusBench v0.2, a chemistry dataset of nearly 50,000 questions, plus post-training recipes. Both are available on Hugging Face. Argonne National Laboratory is integrating Nemotron into a multi-agent system for protein-based therapeutic candidates.
Key Points
- Terray Therapeutics reports its post-trained Nemotron recovered about 76% of annotated reactions on the RxnScribe test.
- Lila Sciences reports lab feedback post-training lifted quantum dot accuracy to 61%, versus 54% for GPT-6 Astra.
- KTH Royal Institute of Technology says its Nemotron agent fixed 81% of GROMACS test cases overnight on one DGX Spark.