Nvidia releases Nemotron 3.5 Super VL and Bolt 74B-A9B/E4B open-weight models on Oct 19, targeting enterprise data centers and on-device agents
Nvidia released three open-weight models for always-on AI agents: 120B Super VL, 74B-A9B Bolt 74B-A9B 24GB, and Bolt E4B to follow later this month; Super VL has been adopted by Skydance, EY, Ramp, CrowdStrike, Perplexity, and Salesforce
On October 19, Nvidia released three open-weight models purpose-built for always-on AI agents and edge scenarios, announced by Kari Briski, Nvidia's VP of open models, in an official blog post.
The flagship Nemotron 3.5 Super VL carries 120 billion parameters and targets enterprise agent workloads. Nvidia says it matches the larger Nemotron 3 Ultra on instruction-following and tool-use accuracy while claiming faster responses than Anthropic Opus 5 and OpenAI's GPT-6.1 Sol, Luna, and Astra on agent tasks including document intelligence, financial accounting, scientific experimentation, medical-insurance claims, and adversarial-defense use cases. The model is aimed at on-premises enterprise data centers, where customers can post-train it on proprietary data.
Nemotron 3.5 Bolt 74B-A9B 24GB has 74 billion parameters and is optimized for on-device multi-step reasoning, tool calling, and web browsing, running on AI PCs with Nvidia RTX Spark silicon and DGX Spark workstations. The smallest Bolt E4B targets conversational agents and is set to ship later this month.
Adoption announcements for Super VL include Skydance, EY, Ramp, CrowdStrike, Perplexity, and Salesforce. Briski's post emphasizes that, in contrast with closed frontier models, enterprises can post-train Super VL on their own data, run it on their own infrastructure, and retain ownership of everything they build. AI PCs carrying Bolt models are reaching market through partners including Microsoft and Dell.