NVIDIA launches Nemotron 3.5 open agent models

NVIDIA has announced Nemotron 3.5 Super VL and two new Nemotron 3.5 Bolt models for open agent deployments across data center, desktop, and edge hardware.

• Super VL is an open 120-billion-parameter model for specialized always-on agents; Bolt 74B-A9B 24GB is available now, while Bolt E4B 4GB arrives on Oct. 23.
• NVIDIA says Super VL is built with the Nemotron Coalition, ships with open weights, open datasets and training recipes, and runs across DGX Spark, DGX Station, RTX PRO, RTX systems, and Jetson; the model is also available via Hugging Face, Open Router, ModelScope, and NVIDIA NIM.
• The post says organizations including Scale AI, AfterQuery, AI21, Mercor, Atlassian, CrowdStrike, Dassault Systèmes, Perplexity, Salesforce, Synopsys, Ramp, Dream, Lila Sciences, R1 and others are already customizing or integrating the models for specialized workflows.