NVIDIAOctober 15, 20263 sources
NVIDIA releases open Nemotron 3.5 Super VL (120B) and Bolt edge models
AI Analysis
Super VL natively handles text, high-resolution document OCR, images and video for enterprise agents. NVIDIA says Dynamo serving on 16 B200 GPUs handles up to 8x more users and cuts first-token latency by 85%.
The Bolt edge checkpoints include a 24GB 74B-A9B model and an upcoming 4GB E4B. NVIDIA also released open datasets alongside the models.
The launch adds to a crowded week of open-weight releases.
