Direct answer: Nvidia has launched new open-weight models built for always-on AI agents — the 120-billion-parameter Nemotron 3.5 Super VL for enterprises, plus two lightweight Bolt models that can run locally on AI PCs. Organisations can post-train them on their own data and run them on their own infrastructure.
What happened
According to TechCrunch, Nvidia launched three models under its Nemotron family. Super VL targets enterprise agentic work — document intelligence, financial accounting, scientific experimentation, insurance claims and security — and Nvidia says it matches the accuracy of its larger Nemotron 3 Ultra on instruction-following and tool use while using fewer parameters.
The Bolt models go the other way: small enough to run locally. Nemotron 3.5 Bolt 74B-A9B 24GB handles multistep reasoning, tool calling and web browsing on AI PCs running the Nvidia RTX Spark chip, while the smaller Bolt E4B is aimed at conversational agents. As Nvidia’s Kari Briski wrote in the announcement: organisations wanting agents on Nvidia infrastructure now get “an open, optimized model for their tier without having to choose between capability and deployability.”
The strategic context matters. Big AI labs are increasingly shopping beyond Nvidia for chips — AWS Trainium, Cerebras, their own silicon — so Nvidia is leaning into the counter-trend: enterprises and consumers who want to run their own models on their own hardware. Local deployment also sidesteps a real worry: privacy-weary users don’t want agents with access to their families, health and finances phoning home to the cloud.
What it means for UK small businesses
Open-weight agent models are arguably the best news for small businesses this quarter. Two blockers keep UK firms from adopting AI agents: per-token API costs that scale unpredictably, and the discomfort of sending customer data to a third party’s servers. A downloadable model you can run on your own machine or a rented workstation kills both at once.
If you’re already exploring agents, our AI agents for small business guide covers the practical setups available today — open models like these slot straight into them. And for a sense of how fast open models are catching frontier labs, see our piece on Vercel’s fastest-adopted open model.
Honest limits / what to watch
- “Faster than Opus 5 and GPT 6.1” is a vendor claim. Nvidia says Super VL beats Anthropic and OpenAI models on post-trained agentic tasks. Treat all vendor benchmarks with scepticism until independent testing confirms them.
- Open weights, not open everything. Nvidia controls the training data and process; you get the weights, not the recipe. “Open” here means deployable, not fully transparent.
- Hardware still matters. The local Bolt models need RTX Spark-powered AI PCs — new hardware you have to buy. For a small business, the cloud APIs may still be cheaper until you hit serious volume.
FAQs
What did Nvidia launch?
Three new open-weight models: Nemotron 3.5 Super VL (120B parameters) for enterprise agents, Nemotron 3.5 Bolt 74B-A9B 24GB for local agentic tasks, and a smaller Bolt E4B for conversational agents.
What are open-weight models?
Models whose trained parameters are publicly released, so organisations can download, post-train on their own data, and deploy them on their own infrastructure instead of relying on a vendor’s cloud API.
Can these models run on a normal PC?
The Bolt models are designed for the new AI PCs running Nvidia’s RTX Spark chip and DGX Spark workstations. Super VL and Bolt 74B are available now; Bolt E4B arrives later this month, according to Nvidia.
Who is already using Nemotron 3.5 Super VL?
Nvidia says users include Skydance, EY, Ramp, Crowdstrike, Perplexity and Salesforce. Nvidia claims faster task completion than rival models from Anthropic and OpenAI on post-trained agentic tasks, though such vendor claims should be treated cautiously.
Why does this matter for UK small businesses?
Open models you can run yourself mean no per-token API bills and no customer data leaving your premises — two of the biggest blockers for small businesses adopting AI agents.
