Nvidia launches Nemotron 3.5 Super VL, a 120B-parameter open model for enterprise agents, and says it outperforms frontier models on specialized benchmarks (Carl Franzen/VentureBeat)

NVIDIA has unveiled three new models in its Nemotron family, aimed at enhancing the performance and efficiency of always-on agents across various devices. The flagship model, Nemotron 3.5 Super VL, boasts 120 billion parameters, delivering frontier-level accuracy for multimodal agentic tasks while being significantly smaller than its predecessors. This model is 5 times faster than the next best open agentic model and can be fine-tuned on an organization's proprietary data, providing unparalleled flexibility and control.

Two additional models, Nemotron 3.5 Bolt 74B-A9B 24GB and Nemotron 3.5 Bolt E4B 4GB, have been introduced to cater to edge devices and local hardware. Bolt 74B-A9B 24GB is designed for everyday agentic tasks, balancing intelligence and speed for optimal performance with 24GB of memory. In contrast, Bolt E4B 4GB is tailored for physical and conversational agents, enabling fast local voice assistants and interactive characters.

These open models cater to a broad range of agent deployments, from large enterprise servers to edge devices and local hardware. NVIDIA's focus on open-source models allows any organization to run agents on NVIDIA infrastructure without compromising on capability or deployability. The release of Super VL and Bolt models marks a significant step forward in AI technology, providing organizations with the tools to build and customize agents tailored to their specific needs.