Nvidia released the open-weight Nemotron 3.5 model series to power always-on artificial intelligence agents, enterprise operations, and local AI personal computers. The new collection features the 120-billion-parameter Nemotron 3.5 Super VL built for enterprise agents, alongside a pair of compact Bolt models engineered to operate locally on workstations and Nvidia-driven personal computers, including fresh RTX Spark laptops.
Nemotron 3.5 Super VL
Nemotron 3.5 Super VL uses a lower parameter count than the earlier Nemotron 3 Ultra while aiming to match its precision regarding instruction following and tool usage operations.
Nvidia stated that Super VL finishes particular agentic duties more rapidly than rival systems such as Claude Opus 5, GPT-6.1 Sol, Luna, and Astra following post-training for tailored tasks.
Those operations encompass financial accounting, document intelligence, medical insurance claims, scientific experiments, and defense against security breaches.
According to Nvidia, organizations like Skydance, Salesforce, EY, Ramp, CrowdStrike, and Perplexity currently use Super VL.
The firm markets this model toward commercial entities seeking to post-train intelligence software on private data and execute it on internal infrastructure rather than relying on closed frontier systems.
Smaller Bolt Models Run Locally
Nvidia additionally revealed a pair of compact systems tailored for local artificial intelligence agents.
Nemotron 3.5 Bolt 74B-A9B 24GB is a 74-billion-parameter variant built for local agentic duties, focusing on operations like web browsing, tool calling, and multistep reasoning while maintaining a balance between logic speed and capability.
The more compact Bolt E4B focuses primarily on conversational intelligence agents.
Such models are built for fresh artificial intelligence personal computers driven by the Nvidia RTX Spark processor and workstations using DGX Spark.
Nvidia noted that operating agents locally permits individuals to execute daily duties without transmitting confidential data concerning health, finances, or family details to cloud servers.
Nvidia Pushes Further Into Open Models
The fresh rollouts broaden the existing Nemotron collection from Nvidia, which currently spans multimodal intelligence, reasoning, mathematics, coding, and business automation.
This rollout arrives as prominent artificial intelligence firms increasingly investigate hardware options outside of Nvidia, encompassing Cerebras and AWS Trainium, alongside creating proprietary microchips.
Simultaneously, Nvidia addresses rising interest from corporations seeking to run open systems within internal data facilities.
Nemotron 3.5 Super VL and Bolt 74B-A9B are accessible currently, and Nvidia reported that Bolt E4B will launch later in the month.
