Nvidia launches open-weight AI models for enterprise and local PC use

Nvidia has launched new open-weight artificial intelligence models for enterprise agents and local computing, with its Super VL and larger Bolt model already available. The company is targeting both organizations operating their own infrastructure and users running AI workloads on PCs.

The enterprise-focused Nemotron 3.5 Super VL has 120 billion parameters. Nvidia said its instruction-following and tool-use accuracy is similar to that of the larger Nemotron 3 Ultra. The company also said that, after training for particular agent tasks, Super VL completes those tasks faster than rival models from Anthropic and OpenAI.

Nvidia cited financial accounting, document intelligence and medical insurance claims among those tasks. It identified Salesforce, EY, Ramp, Skydance, Crowdstrike and Perplexity as users of Super VL. Kari Briski, Nvidia’s vice president of Open Models and Software, said customers can adapt the model with their own data and deploy it using their own infrastructure.

For local workloads, the 74-billion-parameter Nemotron 3.5 Bolt 74B-A9B 24GB is intended to support reasoning across multiple steps, tool calling and web browsing. The smaller Bolt E4B targets conversational agents and will become available later, Nvidia said. The Bolt models are designed for AI PCs using Nvidia RTX Spark and workstations using DGX Spark, with Microsoft and Dell among the PC partners.