Nvidia announced new open‑weight models on Thursday. The Nemotron 3.5 Super VL is a 120‑billion‑parameter model aimed at always‑on enterprise agents. Two lighter Bolt models – a 74‑billion‑parameter version and a smaller E4B variant – are built to run on Nvidia RTX Spark‑powered AI PCs and DGX Spark workstations. Super VL and the 74B Bolt are available now; the E4B Bolt is slated for release later this month.
Nvidia says the Super VL matches the accuracy of its larger Nemotron 3 Ultra while being faster on tasks like document intelligence and tool use, and that the Bolt models let users run agents locally without sending data to the cloud.
Why it matters
Enterprises and knowledge workers can run powerful AI agents on their own hardware, reducing reliance on cloud services.
