Nvidia has released new open-weight models aimed at always-on agents and edge computing. The flagship, Nemotron 3.5 Super VL, is a 120-billion-parameter model for enterprise agent workloads. Nvidia says it uses fewer parameters than the Nemotron 3 Ultra model but delivers similar accuracy on instruction-following and tool-use tasks. The company also introduced two lightweight Bolt models designed to run locally on AI PCs powered by the Nvidia RTX Spark chip and workstations running the DGX Spark. These PCs, aimed at consumers, programmers, and enterprise knowledge workers, are reaching the market through partners such as Microsoft and Dell.

Kari Briski, Nvidia‘s vice president of Open Models and Software, framed the release as giving organizations an open, optimized model for their infrastructure tier without choosing between capability and deployability. Nvidia claims Super VL completes tasks faster than competing models from Anthropic and OpenAI, including Opus 5, GPT 6.1 Sol, Luna, and Astra, when post-trained on specific agentic tasks such as document intelligence, financial accounting, scientific experimentation, medical insurance claims, and security attack resilience. The company backs the claim with a list of customers and partners using Super VL, including Skydance, EY, Ramp, Crowdstrike, Perplexity, and Salesforce.

Super VL joins Nvidia‘s broader Nemotron family, which covers multimodal AI, reasoning, coding, math, and enterprise automation. The announcement comes as fast-growing proprietary AI labs increasingly diversify beyond Nvidia chips, using options like AWS Trainium, Cerebras, and custom silicon. Nvidia is leaning into the trend of enterprises wanting to run their own open models in their own data centers. Briski noted that unlike closed frontier models, organizations can post-train Super VL on their own data, deploy it on their own infrastructure, and own what they build.

With the smaller Bolt models, Nvidia is also pushing the nascent trend of consumer AI assistants that run locally. Privacy-conscious users may be reluctant to give cloud agents access to family, health, and financial information. Running models and agents locally lets PC users delegate everyday projects without sending workloads to the cloud, Briski wrote. Nemotron 3.5 Bolt 74B-A9B 24GB is a 74-billion-parameter model for local agentic tasks, balancing speed with reasoning for multistep reasoning, tool calling, and web browsing. The smallest model, Bolt E4B, targets conversational agents. Super VL and Bolt 74B-A9B are available now, with Bolt E4B expected later this month.