What happened
On 17 November 2026 NVIDIA published a blog post on its SC26 announcements, written around Ian Buck’s keynote. The post introduces Nemotron for Science, describes several science-agent results, and previews a science-focused Rubin GPU, among other items. The figures below are NVIDIA’s, as stated in the post.
Key details
- Nemotron for Science: combines the Nemotron 3.5 Super VL open model with a new open chemistry dataset and post-training recipes for life sciences and chemistry. NVIDIA says it runs locally on a DGX Spark, a workstation or a supercomputer.
- Agent results (NVIDIA): at Argonne, an agent built with the toolkit took one prompt, planned and tested hypotheses, and identified two new binding sites. A KTH agent that repairs broken GROMACS setups fixed 81% of test cases in one overnight run on a single DGX Spark.
- Code porting: NVIDIA says its Nsight AI Porting Agent ported a WRF model routine to GPUs for a speedup of “more than 5x”, using about 150 million tokens.
- Rubin HPC: a science-focused GPU with “more than 50,000 CUDA cores”, about 3x Hopper per the post, spanning FP64 to FP4, expected in 2028. DKRZ selected Vera CPUs for its HLRE-5 supercomputer, expected in early 2028.
- Quantum: NVIDIA says a CUDA-Q decoding algorithm cuts error-correction research computation from 12 weeks to 45 minutes, with a target of reducing decoding latency from about 300 microseconds to 10.
- Photonics: NVIDIA quotes Buck as saying co-packaged optics switches at Lambda are “15x more reliable on bit errors and link flaps”.
- Not stated: prices, licence terms and availability dates for Nemotron for Science, and independent verification of any result.
Why it matters
For researchers, the concrete items are an open model and dataset that can run on a single DGX Spark, and agent examples with measurable outcomes. Those outcomes come from NVIDIA and its partners and have not been independently reproduced, and the Rubin HPC GPU is a roadmap item for 2028 rather than a shipping product.
Disclosure: Claude, made by Anthropic, was one of the AI tools used to research and draft this article. NVIDIA is not affiliated with Deyron Labs; all claims are NVIDIA’s own.
Sources
- Ian Buck Kicks Off NVIDIA's Week at SC26 as Science Enters the Age of Agents Primary source , NVIDIA Blog, 17 November 2026