Nvidia unveils Nemotron 3.5 models for AI agents and PCs — TechCrunch
Nvidia has unveiled new open-weight Nemotron 3.5 models designed for always-on AI agents and edge computing. The lineup includes the enterprise-grade Nemotron 3.5 Super VL with 120 billion parameters, as well as two more compact Bolt models for local deployment on PCs and workstations, TechCrunch reports.
A model for enterprise agents
Nemotron 3.5 Super VL is aimed at use in enterprise AI agents. Nvidia said the model has fewer parameters than Nemotron 3 Ultra but can deliver similar accuracy in following instructions and using tools.
Nvidia also says that, after additional training for specific agent tasks, Super VL performs them faster than a number of competing models. The company named document analysis, financial accounting, scientific experiments, processing insurance claims in healthcare, and resilience to cyberattacks among such tasks. According to Nvidia, Super VL is used by Skydance, EY, Ramp, CrowdStrike, Perplexity, and Salesforce.
More current news is available on the UA.News Telegram channel Telegram.
Local operation on PCs
The two Bolt models are designed to run on new AI PCs with the Nvidia RTX Spark chip and on DGX Spark workstations. Such computers, aimed at consumers, programmers, and corporate knowledge workers, are entering the market through partners including Microsoft and Dell.
Nemotron 3.5 Bolt 74B-A9B 24GB has 74 billion parameters and is designed for local agent tasks requiring multi-step reasoning, tool calls, and web browsing. The smallest model, Bolt E4B, is intended for conversational agents. Super VL and Bolt 74B-A9B are already available, while Nvidia plans to release Bolt E4B later this month.
Nvidia Vice President of Open Models and Software Kari Briski said that organizations can further train Super VL on their own data, deploy it on their own infrastructure, and control the solutions they create. Nvidia also positions local models as a way to delegate everyday tasks without moving workloads to the cloud.
