Nvidia launches Nemotron for Science to put AI agents in the lab Nvidia unveiled Nemotron for Science at the SC26 conference on November 16, 2026, releasing the multimodal Nemotron 3.5 Super VL model and an open chemistry dataset of nearly 50,000 reasoning questions on Hugging Face for public use. Nvidia reported that Lila Sciences reached 61% accuracy in quantum-dot synthesis with Nemotron versus 54% for GPT-6 Astra, at inference costs 10-20 times lower, while a KTH Royal Institute agent fixed 81% of broken GROMACS setups in a single overnight run. The initiative targets chemistry, life sciences, protein design and materials science, with Argonne National Laboratory and Japan's RIKEN among active users. Nvidia launches Nemotron for Science to put AI agents in the lab The open model and chemistry dataset target research in chemistry, life sciences, protein design and materials science Nvidia https://cryptobriefing.com/markets/nvidia/ wants AI agents doing more than writing emails. The chipmaker has introduced Nemotron for Science, an initiative built to speed up scientific discovery with specialized AI agents. The company unveiled the program on November 16, 2026, at the SC26 conference. The package includes a new multimodal model, Nemotron 3.5 Super VL, and an open chemistry dataset of nearly 50,000 reasoning questions. Nvidia’s pitch is that researchers can build domain-specific agents that are more accurate and cheaper to run than general-purpose rivals. What Nvidia actually shipped The core of the release is Nemotron 3.5 Super VL. A multimodal model can work with more than one kind of input, so it isn’t limited to reading plain text. Alongside the model comes the chemistry dataset. Its nearly 50,000 questions are framed as reasoning problems, not simple fact lookups. Nvidia is aiming the tools at four areas: chemistry, life sciences, protein design and materials science. The models and benchmarks are available on Hugging Face for public use. The early results Lila Sciences reached 61% accuracy in quantum-dot synthesis using Nemotron. A competing model, GPT-6 Astra, scored 54% on the same work. AI, tech, and the markets they move—in one daily briefing. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. Inference costs for the Nemotron-based agents were reported to be 10-20 times lower than competing models like GPT-6 Astra. The second showcase came from the KTH Royal Institute. An agent built there fixed 81% of broken GROMACS setups in a single overnight run. Institutional interest is already on the board. Active users include the US Department of Energy’s Argonne National Laboratory and Japan’s RIKEN. How this fits Nvidia’s playbook Nemotron for Science extends Nvidia’s habit of releasing open models paired with substantial datasets and guidance for post-training, the stage where a model is refined for specific tasks. The initiative also builds on earlier work like BioNeMo, Nvidia’s life sciences project. It plugs into the company’s CUDA-X accelerated scientific computing suite as well. That integration is designed to make deployment possible across a range of computing environments. The release also lands within Nvidia’s broader push into agentic AI. Agentic AI refers to systems that can take a sequence of actions toward a goal, not just answer a single prompt. What this means for researchers and the AI race The most immediate stakeholders are labs with tight budgets and long queues. If the reported 10-20 times lower inference cost holds up across real workloads, AI-assisted research becomes affordable for groups that previously couldn’t justify it. The open release on Hugging Face reinforces the point. Public benchmarks also let outside researchers check Nvidia’s claims. The quantum-dot result is a single task, and one benchmark win does not settle which model is better for science broadly. There is also the question of how well these agents generalize. Fixing 81% of broken GROMACS setups is impressive, but the remaining failures still need a human to step in. Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy https://cryptobriefing.com/editorial-policy/ .