Nvidia Launches Nemotron 3.5 Super VL and Bolt Models

Nvidia announced three new Nemotron 3.5 models targeting agentic AI deployments across enterprise and edge hardware. The flagship Super VL is an open 120-billion-parameter multimodal model built with the Nemotron Coalition, trained on 26 trillion tokens, and claimed to match larger reasoning models at one-fifth the size while producing outputs nearly five times faster. Bolt 74B-A9B fits within 24GB of memory using mixture-of-experts and two-bit quantization for local workstation use. A smaller Bolt E4B 4GB model, launching later in October, targets edge devices including wearables, car dashboards and embedded interfaces.