NVIDIA expands Nemotron with Super VL and memory-focused Bolt models

NVIDIA says its 120-billion-parameter Nemotron 3.5 Super VL and Bolt 74B-A9B 24GB are available, with Bolt E4B 4GB scheduled for Oct. 23 (NVIDIA).

NVIDIA says Super VL includes open weights, datasets and training recipes, allowing organizations to customize and host agents themselves. That offers a route to greater deployment control, not proof of privacy or task reliability (NVIDIA).

NVIDIA says Bolt 74B-A9B uses native 2-bit quantization to fit in 24GB on targeted NVIDIA hardware. That memory-focused design may make local agents more practical, but application headroom and workload-specific quality remain unverified. NVIDIA’s claimed Super VL throughput advantage is up to 5x against models in its class, not a universal speed guarantee; comparative performance and production reliability remain unverified (NVIDIA).