Nvidia Releases Nemotron 3.5 Open-Weight AI Models
Nvidia launched three open-weight AI models, including the Nemotron 3.5 Super VL, to enable high-speed multimodal reasoning for enterprise agents.
Nvidia released three new open-weight AI models on October 15, led by the Nemotron 3.5 Super VL. This 120-billion-parameter model is designed for always-on agents capable of reasoning across text, images, documents, and video. It features a hybrid Mamba-Transformer mixture-of-experts architecture and a 1 million-token context window.
Available via Together AI, the Super VL is described by the company as "5x faster than the next best" open agentic model in its class. The evaluation firm Artificial Analysis assigned the model an Intelligence Index score of 26.5, noting its high accuracy for multimodal agentic tasks.
Nvidia also introduced two edge models under the Nemotron 3.5 Bolt umbrella: a 74B-A9B 24GB model for local agentic workloads and an E4B 4GB model for fast conversational agents. These Bolt models allow enterprises to post-train AI on proprietary data and deploy it on local infrastructure to improve privacy. The release aligns with a broader trend of U.S. laboratories challenging the dominance of Chinese open-weight models.
