NVIDIA unveils Nemotron for Science at SC26: an agent at Argonne planned its own hypotheses and found two new binding sites
Synopsis
At SC26, NVIDIA announced Nemotron for Science, an open toolkit for science agents that combines the Nemotron 3.5 Super VL open model with a new open chemistry dataset and post-training recipes for life sciences and chemistry, runnable locally on a DGX Spark, a workstation or a researcher's own supercomputer. At Argonne National Laboratory, an agent built with the toolkit took a single prompt, planned and tested hypotheses, chose which simulations to run and identified two new binding sites. NVIDIA also announced the Rubin HPC GPU, an expanded Vera CPU platform and a quantum computing collaboration with IBM and Fermilab.
Interpretation
NVIDIA released Nemotron for Science, an open model toolkit for running science agents on researchers' own systems, comprising the Nemotron 3.5 Super VL open model, a new open chemistry dataset and post-training recipes for life sciences and chemistry. Previously this kind of agent did not carry the interaction with supercomputers, instruments and devices; NVIDIA high-performance computing head Ian Buck said agents have changed how researchers interface with supercomputers, devices and instruments and how they do research. At Argonne National Laboratory, an agent built with the toolkit took a single prompt, planned and tested hypotheses on its own, chose which simulations to run and identified two new binding sites.
Agents call tools between model steps and lean on CPUs to run, evaluate and gather results in that loop. NVIDIA cites this loop as the reason for an expanded Vera CPU platform, and Germany's climate computing center DKRZ has selected Vera CPUs for its HLRE-5 supercomputer, expected in early 2028. Buck said Vera systems were already announced at the Texas Advanced Computing Center, with DKRZ the latest site to join.
Perspective
The report covers what NVIDIA announced in its SC26 special address: the composition and local-running mode of Nemotron for Science, agent use cases at Argonne, KTH and other institutions, the Rubin HPC GPU and Vera CPU platform, and the planned quantum computing collaboration with IBM and Fermilab. Rubin HPC is expected in 2028, DKRZ's HLRE-5 in early 2028, and the quantum collaboration was announced as plans.
The Argonne binding-site result and KTH's 81% fix rate are cases relayed by the announcing company in the address, and the material gives no comparison baseline or evaluation detail; Rubin HPC and DKRZ's HLRE-5 are both future time points, and the quantum error-correction and decoding figures likewise come from statements in the address.
