NVIDIA's Nemotron 3.5 Family Goes Open — and Goes Small

NVIDIA released Nemotron 3.5 Super VL, a 120-billion-parameter open model it claims delivers the fastest task completion and highest accuracy for specialized always-on agents. Two edge siblings joined it: Nemotron 3.5 Bolt 74B-A9B with a 24GB footprint, tuned for local agentic workloads, and Nemotron 3.5 Bolt E4B at just 4GB. The release extends NVIDIA's push to make its model family the default substrate for agents that run continuously rather than in bursts.

Super VL is a 120B open-weight model positioned around accuracy and speed for persistent agent workloads

Bolt 74B-A9B fits in 24GB, optimizing intelligence per gigabyte for local agentic use

Bolt E4B shrinks to a 4GB footprint, targeting constrained devices and edge deployments

Source: https://blogs.nvidia.com/blog/nemotron-3-5-super-vl/