Nvidia releases Nemotron open models spanning language, robotics and biology

By Splitfeed EditorialLast updated Oct 21, 2026, 10:01 UTC

On October 21, 2026, Nvidia announced an expansion of its Nemotron open model portfolio through a post on the official Nvidia blog. The release introduced two new models, Nemotron 3.5 Super VL and Nemotron 3.5 Bolt, alongside Nemotron Retriever models designed for embedding and reranking tasks. The portfolio now spans four domains: language, robotics, biology, and quantum computing. Nvidia framed the announcement explicitly around sovereign AI development, emphasizing support for local languages and regional industries, with Europe named as a primary target audience.

The sovereign AI framing is the detail that lifts this beyond a routine model drop. Sovereign AI refers to the capacity of countries or regions to develop and control AI systems in their own languages and on their own terms, rather than depending on infrastructure and models produced elsewhere. By positioning Nemotron models as tools for that goal, Nvidia is competing not just on technical capability but on political and economic relevance to European governments and enterprises weighing dependency on US-based AI providers. What remains genuinely in dispute is whether open model releases of this kind meaningfully transfer capability and control, or whether they leave the underlying compute and ecosystem dependency with Nvidia regardless.

With no published reactions from any camp at this stage, it is possible to sketch what each would typically argue. Pro-AI voices would likely welcome the open release as democratizing access to capable models across high-value domains, reducing barriers for smaller organizations and non-English-speaking regions. Anti-AI voices would probably raise concerns about the concentration of foundational AI infrastructure in a single large vendor, questioning whether framing commercial releases as sovereign tools obscures continued dependency. The middle ground camp would likely focus on the practical conditions that determine whether open models genuinely serve sovereign goals, including whether local compute, expertise, and governance structures exist to support them.

The argument about what these models can actually do in practice will sharpen once independent benchmarks are published and early adopters in European research or government contexts report results. Any formal partnerships with European institutions or regulatory responses to the sovereign AI framing would also clarify how much traction Nvidia's positioning gains beyond the announcement itself.