Nvidia on Thursday launched new open-weight models for always-on AI agents, aimed at corporate data centers and AI personal computers. The headline model, Nemotron 3.5 Super VL, has 120 billion parameters and is available now, while two lightweight Bolt models run locally on Nvidia-powered PCs.

Nvidia on Thursday launched new open-weight models built for always-on agents, targeting corporate data centers and a new class of AI personal computers that are reaching the market now.

What Are Nvidia's New Open Models for AI Agents?

The headline model is Nemotron 3.5 Super VL, a large system with 120 billion parameters aimed at enterprise agents. Nvidia says it uses fewer parameters than its Nemotron 3 Ultra model but can deliver similar accuracy on instruction-following and tool-use tasks.

Can AI PCs Run AI Agents Locally?

At the other end of the spectrum, the company introduced two lightweight Bolt models meant to run locally. They are designed for AI PCs built on the Nvidia RTX Spark chip and for workstations running DGX Spark. Those PCs, aimed at consumers, programmers and enterprise knowledge workers, are being released through partners including Microsoft and Dell.

“Any organization that wants to run agents on Nvidia infrastructure now has an open, optimized model for their tier without having to choose between capability and deployability,” Kari Briski, Nvidia’s vice president of Open Models and Software, wrote in the blog post announcing the models.

Nvidia claims Super VL completes tasks faster than competing models from Anthropic and OpenAI, naming Opus 5, GPT 6.1 Sol, Luna and Astra, when it is post-trained on specific agentic tasks. Those tasks include document intelligence, financial accounting, scientific experimentation, medical insurance claims and security attack resilience.

Which Companies Are Using Nvidia's Super VL Model?

Performance claims are standard whenever a major tech company unveils a new model. Nvidia has also published a list of companies it says are already using Super VL, including Skydance, EY, Ramp, Crowdstrike, Perplexity and Salesforce.

The release is not the first time Nvidia has put open-weight models up against proprietary frontier labs. Those labs are both its customers and its investments. The company already maintains an extensive Nemotron family that covers multimodal artificial intelligence, reasoning, coding, math and enterprise automation.

Those fast-growing proprietary labs are also looking beyond Nvidia for computing power. They are turning to chips from AWS, in the form of Trainium, and from Cerebras, as well as building their own chips.

Nvidia is betting that more enterprises will want to run their own open models in their own data centers. “Unlike closed frontier models, organizations can post-train Super VL on their own data, deploy it on their own infrastructure and own what they build,” Briski wrote.

The smaller models also point to consumer AI assistants. New assistants appear almost every week from startups and Big Tech alike, but privacy-conscious consumers may hesitate to give those agents access to information about their families, health and finances. Nvidia is pitching local processing as the answer. Briski wrote that PC users “can delegate everyday projects … without sending those workloads to the cloud.” That approach also gives Nvidia a reason for people to buy PCs built around its chips.

Of the two Bolt models, Nemotron 3.5 Bolt 74B-A9B 24GB has 74 billion parameters and is designed for local agentic tasks. Nvidia says it balances speed with reasoning to handle multistep reasoning, tool calling and web browsing. The smallest model, Bolt E4B, is aimed at conversational agents.

Super VL and Bolt 74B-A9B are available now. Bolt E4B is coming later this month, Nvidia said.

This article was written with the assistance of AI.
News Factory APP - agentic news to boost your SEO & AEO.