Nvidia launches Nemotron 3.5, open models for companies and PCs
Open agent models in several sizes, some built to run on a local Nvidia machine.
Nvidia has released new open-weight Nemotron 3.5 models. Nemotron 3.5 Super VL is a 120-billion-parameter model for business agents, which Nvidia says delivers accuracy similar to the larger Nemotron 3 Ultra on instruction following and tool use. Two smaller Bolt models are meant to run locally on new AI PCs with Nvidia's RTX Spark chip and on DGX Spark workstations.
Open weights mean companies can train the model further on their own data and run it on their own servers, which Nvidia presents as the main difference from closed models. The Bolt models let PC users hand off everyday work without sending it to the cloud, according to Nvidia. Its speed claims against Anthropic and OpenAI models are its own.
Super VL and Nemotron 3.5 Bolt 74B-A9B are available now. Bolt 74B-A9B is built for local agent work such as multistep reasoning, tool calling and web browsing. The smallest model, Bolt E4B, is aimed at conversational agents and was released on 23 October, according to Nvidia.