Nvidia has launched a new set of open weight models aimed at always on agents. Nemotron 3.5 Super VL is a 120 billion parameter model for enterprise agents, which Nvidia says delivers accuracy similar to its larger Nemotron 3 Ultra on instruction following and tool use tasks, and can be post trained on a customer’s own data and deployed on its own infrastructure.
The company named Skydance, EY, Ramp, CrowdStrike, Perplexity and Salesforce as users of Super VL, and said the model is aimed at tasks including document intelligence, financial accounting, scientific experimentation, medical insurance claims and security work. Those performance and adoption claims are Nvidia’s own.
At the small end, two Bolt models are built to run locally. Bolt 74B-A9B 24GB is a 74 billion parameter model for multistep reasoning, tool calling and web browsing on the new RTX Spark AI personal computers and DGX Spark workstations sold through partners including Microsoft and Dell. A smaller Bolt E4B, for conversational agents, follows later this month. Super VL and the larger Bolt are available now.
“Any organization that wants to run agents on Nvidia infrastructure now has an open, optimized model for their tier without having to choose between capability and deployability,” Kari Briski, Nvidia’s vice president for open models and software, wrote. The local models also serve a privacy pitch: agents that run on the device do not send family, health or financial workloads to the cloud.