Nvidia releases open models for enterprise agents and local AI PCs
Organisations can run these models on their own infrastructure and data rather than depending only on closed frontier APIs.
Nvidia has released a set of new open-weight models aimed at agents running in enterprises and on local hardware. The Nemotron 3.5 Super VL, a 120-billion-parameter model, is pitched for enterprise agent workloads, with Nvidia claiming accuracy comparable to the larger Nemotron 3 Ultra on instruction-following and tool-use tasks. Two lightweight Bolt models target AI PCs running the RTX Spark chip and workstations with DGX Spark, with the 74-billion-parameter Bolt 74B-A9B available now and the smaller Bolt E4B due later this month. For developers and companies, the pitch is that they can post-train and host these models on their own data and infrastructure instead of relying on closed frontier APIs, and that local models let consumers run assistants without sending personal data to the cloud. Nvidia's broader open-model push also includes Kumo structured-data models, Cosmos world models, and Agent Knowledge Foundation tooling that enterprise partners such as Palantir, Databricks and Snowflake are already integrating.
Why it matters
The release gives companies a route to run capable agents on their own infrastructure and keeps sensitive company data in-house instead of sending it to outside providers. For consumers, local models mean an assistant can run on the device without personal data leaving it. Taken together, the launch makes open, self-hosted models a more practical choice for organisations that have been relying on closed services, and extends that choice to individual computers.
Matters or hype?
Does this story matter, or is it hype? Vote before you see what everyone else said.
Sources · 4
- NVIDIA Blog ↗ Primary source
- NVIDIA Blog ↗ Primary source
- NVIDIA Blog ↗ Primary source
- TechCrunch ↗