Nvidia Unveils Nemotron 3.5 Super VL and Nano Open AI Models
Nvidia announced two open AI models on October 15, 2026: the 120-billion-parameter Nemotron 3.5 Super VL and the smaller Nemotron 3.5 Nano.
Nvidia unveiled the open AI models Nemotron 3.5 Super VL and Nemotron 3.5 Nano on October 15, 2026, according to digitaltoday.co.kr.
Nemotron 3.5 Super VL has 120 billion parameters and is suited for always-on AI agent development. Nvidia said it delivers performance similar to its largest inference model at about one-fifth the size, and handles text as well as images and video.
Atlassian, CrowdStrike, Dassault Systèmes, Salesforce and Synopsys fine-tuned Nemotron 3.5 Super VL on their own data, according to the company.
Nemotron 3.5 Nano, listed as Nano 74B-A9B 24GB, is a smaller model based on Super VL that runs agent tasks on-device with 74 billion parameters and a hybrid mixture-of-experts structure activating 9 billion parameters at a time. It runs on enterprise workstation or PC GPUs, handling tasks like coding and debugging locally without sending them to the cloud.
Nvidia will release a smaller Nano E4B 4GB model at the end of October, targeting edge devices such as GeForce RTX, Jetson and RTX/GTX Spark for wearables, home devices, point-of-sale terminals, car dashboards and embedded interfaces.
Nvidia's open-model push
According to the Wall Street Journal, the move is seen as continuing innovation in open-weight and open-source models and could lend momentum to the US open model camp. Kari Briski, Nvidia vice president of enterprise generative AI software, said public institutions and financial institutions are reluctant to use Chinese-made models, and that interest in Nemotron is growing because it is American-made and its training data can be examined.
Quick answers
When did Nvidia announce Nemotron 3.5 Super VL and Nemotron 3.5 Nano?
Nvidia unveiled both open models on October 15, 2026.
What are the key specs of the two models?
Nemotron 3.5 Super VL has 120 billion parameters and handles text, images and video. Nemotron 3.5 Nano has 74 billion parameters with a hybrid mixture-of-experts structure activating 9 billion parameters at a time.
When does the smaller Nano E4B 4GB model arrive?
Nvidia will release the Nano E4B 4GB model at the end of October, targeting edge devices such as GeForce RTX, Jetson and RTX/GTX Spark.