Nvidia releases open models for enterprise agents and local PCs

Nvidia on Thursday launched open-weight AI models for enterprise agents and devices that run agents locally. Nemotron 3.5 Super VL is available now, alongside one of two new lightweight Bolt models.

Super VL has 120 billion parameters and targets enterprise agents. It is smaller than Nvidia’s Nemotron 3 Ultra, but Nvidia says it delivers similar accuracy on instruction-following and tool-use tasks.

Nvidia also claims Super VL completes tasks faster than competing models from Anthropic and OpenAI, including Opus 5, GPT 6.1 Sol, Luna and Astra, when post-trained for specific agent tasks. The tasks Nvidia cited cover document intelligence, financial accounting, scientific experimentation, medical insurance claims and resilience against security attacks. Nvidia says Skydance, EY, Ramp, Crowdstrike, Perplexity and Salesforce are using Super VL.

The release expands Nvidia’s Nemotron family, which already includes models for multimodal AI, reasoning, coding, math and enterprise automation. Nvidia competes with proprietary AI labs on models while also supplying and investing in them. Those labs are looking to other chip and AI system providers, including AWS’s Trainium and Cerebras, and developing chips of their own.

Businesses can run Super VL on their own systems

Nvidia is pitching Super VL to companies that want to operate open models in their own data centers. Organizations can post-train it on their data and deploy it on their infrastructure, Nvidia vice president of Open Models and Software Kari Briski wrote in the announcement.

The Bolt models address local use on AI PCs with Nvidia’s RTX Spark chip and workstations running DGX Spark. Partners including Microsoft and Dell are bringing the PCs to market for consumers, programmers and enterprise knowledge workers.

Local processing is also part of Nvidia’s pitch for consumer assistants. An assistant with access to family, health or financial information may raise privacy concerns. Briski wrote that PC users can hand off everyday projects “without sending those workloads to the cloud.”

Nemotron 3.5 Bolt 74B-A9B 24GB has 74 billion parameters. Nvidia designed it for local agent tasks that involve multistep reasoning, tool calls and web browsing, while balancing speed and reasoning ability. The smaller Bolt E4B is aimed at conversational agents.

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