Nvidia releases Nemotron models for enterprise and local AI agents
Nvidia released Super VL and Bolt models for AI agents running on company infrastructure and local devices, with another version due October 23.
NVIDIANvidia announced additions to its Nemotron model family on October 15 for enterprise systems and personal computers. Super VL and one Bolt version are available, while a smaller Bolt model is scheduled for October 23.
Nemotron 3.5 Super VL has 120 billion parameters and combines visual understanding with text processing. Nvidia says it can interpret documents, charts and web pages while supporting tasks that require multiple steps.
The Bolt 74B-A9B model uses two-bit quantization to fit into 24GB of memory. Nvidia designed it for local workloads such as research and coding on compatible hardware.
The releases let organizations customize models using their own data and operate them on infrastructure they control. Nvidia is also providing model weights, datasets and training methods.
The forthcoming Bolt E4B targets a 4GB memory footprint for conversational and physical agents. Nvidia lists voice assistants and embedded interfaces among its intended uses..
The day's biggest AI news in one short email, every morning. Plus an instant heads-up when a reset or a big release lands.
More on NVIDIA
- Oct 19Nvidia expands broadcaster AI framework and Adobe video search collaborationOfficial
- Oct 19Tenstorrent raises $1.5 billion at a $6.2 billion valuationNews
- Oct 19Germany and Nvidia agree to expand industrial AI cooperationOfficial
- Oct 19Nscale and Anthropic reportedly discuss second $45 billion computing dealStrong rumor
- Oct 20Barcelona AI accelerator expands startup support and Nvidia accessOfficial
- Oct 20ByteDance, Alibaba and Tencent reportedly order Nvidia cards worth over $35 billionNews