Nvidia has launched three open-weight Nemotron 3.5 AI models designed for enterprise automation, always-on AI agents, and local computing. The lineup includes a 120-billion-parameter model for complex business tasks and two smaller models that can run directly on AI-powered PCs and workstations without relying on cloud processing.
The models, announced on October 15, are intended to support organizations running AI agents on Nvidia infrastructure. They also allow developers to customize models using their own data and deploy them within privately managed computing environments.
Nemotron 3.5 Super VL Targets Enterprise AI Tasks
The largest model, Nemotron 3.5 Super VL, contains 120 billion parameters and supports instruction-following, tool use, visual understanding, and multistep business workflows. Nvidia claims it delivers accuracy comparable to its larger Nemotron 3 Ultra model while using approximately one-fifth as many parameters.
Nvidia also claims Super VL can complete specialized agent tasks up to five times faster than competing models in its class. Its reported performance includes applications involving document analysis, financial accounting, scientific research, medical insurance claims, and cybersecurity.
According to Nvidia, companies using Super VL include Skydance, EY, Ramp, CrowdStrike, Perplexity, and Salesforce. Kari Briski, Nvidia’s vice president of Open Models and Software, said organizations can further train the model using proprietary data and operate it on their own infrastructure.
The model joins Nvidia’s existing Nemotron family, which supports reasoning, coding, mathematics, multimodal processing, and enterprise automation.
Smaller Bolt Models Run AI Agents Locally
Nvidia also introduced Nemotron 3.5 Bolt 74B-A9B 24GB, a 74-billion-parameter model designed for local AI applications. It uses 2-bit quantization to operate within 24GB of memory while handling multistep reasoning, tool execution, web browsing, and coding tasks.
The model supports Nvidia RTX Spark-powered laptops, DGX Spark systems, GeForce RTX PCs, and RTX PRO workstations. These devices allow users to perform AI-assisted tasks locally without sending sensitive information to external cloud services.
The smaller Nemotron 3.5 Bolt E4B 4GB targets conversational and physical AI applications, including voice assistants, interactive characters, and embedded interfaces. It supports Nvidia GeForce RTX, Jetson, and Spark hardware.
Nvidia said the models are intended for consumers, developers, and enterprise employees, with new RTX Spark-powered computers reaching the market through manufacturers including Microsoft and Dell.
The company is also promoting open-weight models as AI developers increasingly consider alternative computing infrastructure, including AWS Trainium, Cerebras, and custom-designed processors.
Nemotron 3.5 Super VL and Bolt 74B-A9B are currently available, with support for development tools including Hugging Face, Ollama, LM Studio, and vLLM. Nvidia confirmed that Bolt E4B will become available on October 23, 2026.
Featured image credits: Wikimedia Commons
For more stories like it, click the +Follow button at the top of this page to follow us.