The latest in AI, every dayAI News
← AI News

October 19, 2026 · TechCrunch

Nvidia launches new open models to run agents in the enterprise and on PCs

My take: Nvidia just launched its Nemotron 3.5 family: a 120-billion-parameter model built for enterprises to run their own AI agents on their own infrastructure, plus smaller models so anyone with an RTX-equipped PC can run agents locally, without sending anything to the cloud. That's a real shift for anyone handling information they don't want (or aren't allowed) to upload to someone else's server.

Nvidia says its enterprise model solves agent tasks, like medical insurance claims or financial compliance, faster than models from Anthropic and OpenAI. Since that comparison was run by the same company that makes the product, it's worth reading those numbers with your own critical lens and waiting for independent testing before treating them as definitive.

For a business, this opens the door to training a model on your own data and running it on your own cloud, without that information leaving your control. For someone who uses their computer to work with AI every day, it means being able to delegate multistep tasks (searching, organizing, resolving) without depending on a constant connection or paying for every cloud query.

If you handle sensitive data at work (client, patient, or financial information), how comfortable are you today sending it to a cloud AI model? And if you own a business, have you considered running your own models instead of depending entirely on an outside provider?

Read at the source: TechCrunch ↗

Want to use these tools? See the unbiased reviews or back to the news.