Nvidia introduces open-weight AI models for enterprise and consumer use

Nvidia has launched two new open-weight AI models, the Nemotron 3.5 Super VL and Bolt series, designed for enterprise and consumer applications. The Super VL, a 120-billion-parameter model, is optimized for enterprise agents and edge computing, offering comparable accuracy to its larger counterpart, the Nemotron 3 Ultra, despite having fewer parameters. The Bolt models, including the 74-billion-parameter Bolt 74B-A9B 24GB and the smaller Bolt E4B, are intended for local execution on Nvidia’s new AI PCs and workstations, targeting consumers, programmers, and enterprise users. Nvidia’s vice president, Kari Briski, emphasized that these models allow organizations to train and deploy agents on their own infrastructure, enhancing control and privacy. The Super VL is claimed to outperform competing models from Anthropic and OpenAI in specific agentic tasks, with major clients like Skydance, EY, and Salesforce already adopting it. The Bolt models aim to support local AI assistants, addressing consumer concerns about privacy by keeping data on-device rather than in the cloud. The release coincides with a broader industry shift toward open models, as companies seek to reduce reliance on proprietary AI labs. Nvidia is positioning itself as a key player in this trend, leveraging its hardware and software ecosystem to drive adoption in both enterprise and consumer markets.