Nvidia Launches Nemotron 3.5 Models for Enterprise and PC Agents

Disclosure: this post contains affiliate links. If you buy through them we may earn a commission at no extra cost to you. Learn more.

TL;DR: Nvidia launched new open-weight Nemotron 3.5 models designed for always-on agents and edge computing, including the 120-billion-parameter Super VL for enterprise use and lightweight Bolt models for AI PCs. The models allow organizations to run agents on their own infrastructure without relying on cloud services, with companies like Salesforce, Perplexity, and Crowdstrike already using Super VL.

What's New

Nvidia announced the release of new open-weight models in its Nemotron 3.5 family on Thursday, October 19, 2026. The lineup includes models designed for both enterprise deployments and local AI PCs.

The flagship Nemotron 3.5 Super VL is a 120-billion-parameter model built for enterprise agents. Despite using fewer parameters than the Nemotron 3 Ultra model, Nvidia claims it delivers similar accuracy in instruction-following and tool-use tasks. According to Nvidia's vice president of Open Models and Software, Kari Briski, the 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 on specific agentic tasks.

These specialized tasks span document intelligence, financial accounting, scientific experimentation, medical insurance claims, and security attack resilience. Nvidia reports that several major companies are already using Super VL, including Skydance, EY, Ramp, Crowdstrike, Perplexity, and Salesforce.

On the consumer side, Nvidia introduced two lightweight Bolt models designed to run locally on AI PCs equipped with the Nvidia RTX Spark chip and workstations running the DGX Spark. The Nemotron 3.5 Bolt 74B-A9B 24GB is a 74-billion parameter model designed for local agentic tasks, balancing speed with reasoning to handle multistep reasoning, tool calling, and web browsing. The smaller Bolt E4B targets conversational agents.

"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," Briski wrote in the announcement blog post.

Pricing and Availability

Nvidia has not disclosed pricing information for the new models. Super VL and Bolt 74B-A9B are available now, while Bolt E4B is scheduled to arrive later this month, according to the announcement.

The Bolt models are designed to run on new AI PCs featuring Nvidia RTX Spark chips, which are being released through partners including Microsoft and Dell. These PCs target consumers, programmers, and enterprise knowledge workers.

How to Try It

Specific download or access instructions were not provided in the announcement. As open-weight models, organizations should be able to deploy them on their own infrastructure. The models are part of Nvidia's broader Nemotron family, which includes models for multimodal artificial intelligence, reasoning, coding, math, and enterprise automation.

Unlike closed frontier models, organizations can post-train Super VL on their own data and deploy it on their own infrastructure, according to Briski. This approach allows companies to maintain ownership of what they build.

Who It's For

The Nemotron 3.5 family serves two distinct audiences. Super VL targets enterprise organizations that want to run their own open models in their own data centers rather than relying on cloud-based proprietary models. This addresses the growing trend of companies seeking greater control over their AI infrastructure and data.

The Bolt models aim at a different market: privacy-conscious consumers and knowledge workers who want AI assistants running locally on their PCs. By processing workloads locally rather than in the cloud, users can delegate everyday projects involving sensitive information about families, health, and finances without sending data externally.

This dual approach reflects Nvidia's strategy to serve both enterprise customers and create a new market for AI-powered PCs running its chips.

Alternatives

Nvidia's announcement positions these models against proprietary offerings from Anthropic and OpenAI, including their Opus 5, GPT 6.1 Sol, Luna, and Astra models. The company notes that these frontier labs—which are both Nvidia customers and investment targets—are increasingly looking to alternative chip providers like AWS's Trainium and Cerebras, as well as developing their own custom chips.

For organizations exploring AI tools, Perplexity offers AI-powered search capabilities, while platforms like Replit provide development environments that integrate AI assistance. However, these cloud-based tools differ from Nvidia's local deployment approach.