Nvidia launches new open models to run agents in the enterprise and on PCs
TechCrunch Julie Bort
Nvidia launched new open AI models for enterprise agents and for PCs. It wants businesses and consumers running more AI locally, not just in the cloud.
Based on reporting by TechCrunch, Julie Bort — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Nvidia is pushing two ends of the AI market at once. On Thursday, it released open-weight models for always-on agents in the enterprise and lighter models meant to run on PCs and workstations at the edge.
The headline model is Nemotron 3.5 Super VL, a 120-billion-parameter system aimed at enterprise agents. Nvidia says it can match the Nemotron 3 Ultra on instruction-following and tool-use tasks, even though it uses fewer parameters. The company says it also runs faster than competing models from Anthropic and OpenAI — Opus 5, GPT 6.1 Sol, Luna and Astra — after post-training on specific agentic jobs.
Those jobs are not abstract benchmarks. Nvidia points to document intelligence, financial accounting, scientific experimentation, medical insurance claims and security attack resilience. It also says companies including Skydance, EY, Ramp, Crowdstrike, Perplexity and Salesforce are using Super VL. That is the kind of customer list Nvidia likes to flash when it wants to show the model is already out in the wild.
Then there are the small models, and they tell the other half of the story. Nemotron 3.5 Bolt 74B-A9B 24GB is built for local agentic tasks, with a focus on speed, reasoning, tool calling and web browsing. The smaller Bolt E4B is for conversational agents. Super VL and Bolt 74B-A9B are available now, while Bolt E4B lands later this month.
The bigger pitch is control. Nvidia is arguing that companies should be able to post-train models on their own data and run them on their own infrastructure, instead of sending work to closed frontier systems. At the same time, it is trying to make a case for consumer AI assistants that stay on-device, on new AI PCs powered by RTX Spark chips and DGX Spark workstations, sold through partners like Microsoft and Dell. Privacy-conscious users may like the sound of that. Nvidia definitely does.
My take — AI-written commentary, not fact-checked reporting
This is Nvidia doing what Nvidia does best: selling the picks and shovels, then selling the mine cart, then selling the road into the mine. Open models are the useful part here, not the marketing poetry about “agents.” The real prize is simpler: if customers want AI that stays inside their own walls, Nvidia wants to be the default hardware and model stack for the job.
Read more about this at: TechCrunch