Nvidia has released a new set of open-weight Nemotron 3.5 models built to power AI agents, from large enterprise deployments down to consumer PCs. The headline model, Nemotron 3.5 Super VL, targets business automation, while two smaller Bolt models are designed to run locally on PCs with Nvidia's RTX Spark chip. Here is what was announced and what it means.
Nemotron 3.5 Super VL for the enterprise
Super VL has 120 billion parameters, fewer than the earlier Nemotron 3 Ultra, but Nvidia says it matches that model's accuracy on instruction-following and tool-use tasks. After post-training on specific agentic jobs, Nvidia claims it completes tasks faster than competing models from Anthropic and OpenAI. TechCrunch notes Nvidia did not publish benchmark figures to back that up.
Nvidia lists target workloads including document intelligence, financial accounting, scientific experimentation, medical insurance claims and resilience against security attacks. Named early adopters include Skydance, EY, Ramp, CrowdStrike, Perplexity and Salesforce.
Bolt models for AI PCs
- Nemotron 3.5 Bolt 74B-A9B: a 74-billion-parameter model for local agentic tasks, balancing speed and reasoning for multistep reasoning, tool calling and web browsing. Available now.
- Nemotron 3.5 Bolt E4B: the smallest model, aimed at conversational agents. Due later in October.
Both Bolt models run on PCs with Nvidia's RTX Spark chip and on DGX Spark workstations, sold through partners including Microsoft and Dell. Running agents on the device keeps data local instead of sending it to a cloud service, which matters for privacy-conscious users and businesses handling sensitive files.
Why it matters
Open-weight models that companies can tune and host themselves give Nvidia another way to keep developers on its platform, at a time when rivals such as AWS are working to make AI software less dependent on Nvidia hardware. For consumers, the Bolt models are an early test of whether always-on AI assistants that run entirely on a laptop can be genuinely useful.
Should you buy an RTX Spark PC for local AI?
Only if you have a clear local AI workload today. RTX Spark machines are expensive, and early reviews show they excel at running large models on the device but lag in single-threaded speed and gaming. Our guide to running AI models locally explains how much memory and GPU power different models need, and our look at 2026 GPU prices covers why AI hardware costs more this year.
What is Nvidia Nemotron 3.5 Super VL?
It is a 120-billion-parameter open-weight model from Nvidia designed for enterprise AI agents, handling tasks such as document intelligence and financial accounting.
Can Nvidia's Nemotron models run on a PC?
Yes. The Nemotron 3.5 Bolt models are designed to run locally on PCs with Nvidia's RTX Spark chip and on DGX Spark workstations.
When is Nemotron 3.5 Bolt E4B available?
Nvidia says the smaller Bolt E4B model will arrive later in October 2026. Super VL and Bolt 74B-A9B are available now.
