Analysis
Nemotron 3.5 SuperVL and Bolt74B-A9B became available on October 15, NVIDIA says. The smaller Bolt E4B is scheduled for October 23 and remains forthcoming as of October 16, leaving the launch spread across two dates.
NVIDIA’s announcement describes open weights, data and training recipes, alongside reported benchmark and throughput gains. For operators considering local deployment, the accompanying materials broaden what is available to examine beyond the models’ published scores.
The serving calculation still begins with the work a model must do. A reported throughput improvement matters only alongside the quality required from its answers and the hardware needed to deliver them. Expected request volume then changes the economics: a model’s value to a particular service cannot be read directly from a benchmark headline.
Downloadable weights also put operating decisions in the hands of the team running the service. Prompts, retrieved material and outputs each have a path through that system. Hosting the model locally does not by itself establish privacy or security; those properties depend on the surrounding deployment.
The two released models can now be assessed on that basis. E4B remains a future option, with its place in the lineup dependent on the announced release arriving. Its smaller size alone cannot settle whether it will suit a particular workload.
Source: Open From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and Bolt Models · 2026-10-15 · Primary source context · attributed account