Nvidia releases open Nemotron 3.5 Super VL and Bolt models

Nvidia released Nemotron 3.5 Super VL, an open 120-billion-parameter model for AI agents, along with two smaller Bolt models for local hardware, Nvidia’s Kari Briski wrote Thursday in a company blog post.

What Super VL does: Nvidia says the model handles instruction following, tool use, multistep business workflows and visual material such as charts, tables, web pages and PDFs. It says the model’s accuracy is comparable with Nvidia’s largest reasoning model at roughly one-fifth the parameter count, and that it is the fastest open agentic model in its class, “5x faster than the next best,” citing Artificial Analysis. Nvidia said it was trained on about 26 trillion tokens.

For local machines: Nemotron 3.5 Bolt 74B-A9B 24GB uses 2-bit quantization to fit in 24 GB of memory and runs on RTX PCs and DGX Spark, Nvidia said. A smaller Bolt E4B 4GB, for voice assistants and other conversational agents, is coming Oct. 23.

Availability: Super VL and Bolt 74B-A9B are available now on Hugging Face. Super VL ships with open weights, open datasets and training recipes, Nvidia said. Nvidia says companies including Scale AI, Atlassian, CrowdStrike, Perplexity and Salesforce are customizing or integrating Super VL.

Built for: Nvidia says Bolt 74B-A9B is built for its RTX Spark laptops, which it said would arrive Oct. 16.