Nvidia releases open Nemotron 3.5 Super VL and Bolt models

Nvidia released the Nemotron 3.5 Super VL model and two Bolt edge models on October 15, 2026. MarkTechPost, ZDNet AI, and SiliconANGLE reported that the release includes open-weight models designed for artificial intelligence agents and edge systems. [1][2][3]

Nemotron 3.5 Super VL is a 120-billion-parameter open multimodal mixture of experts model that utilizes 12 billion active parameters. MarkTechPost reported that the model features a 1-million-token context window and uses NVFP4 weights to operate on a single DGX Spark system. ZDNet AI reported that Nvidia claims the model is five times faster than the next best open agentic model in its class. [1][2]

Nvidia also announced two 3.5 Bolt edge models to support local hardware and edge systems. SiliconANGLE reported that the total offering of three new models aims to cover agent deployments ranging from enterprise servers to small local devices. [3]

Nemotron is a family of artificial intelligence models created by Nvidia that includes large language models and multimodal systems. The family is designed to support tasks such as reasoning, programming, information retrieval, and agent applications. Nvidia has previously provided open model weights, training data, and training methods for elements of the Nemotron series. [10]

ZDNet AI reported that the new models join a broader wave of open-weight releases from United States developers that compete with open models developed by technology laboratories based in China. [2]