Nvidia released a set of open-weight models on Thursday aimed at always-on agents and edge computing, splitting the lineup between a large enterprise model and two lightweight models built to run locally on AI PCs. The enterprise model, Nemotron 3.5 Super VL, carries 120 billion parameters and is positioned for agents used by businesses, while the two Bolt models are intended for machines running Nvidia's RTX Spark chip and workstations running the DGX Spark.
Nvidia says Super VL, despite using fewer parameters than its Nemotron 3 Ultra model, can deliver similar accuracy on instruction-following and tool-use tasks. The company also claims Super VL finishes tasks faster than competing models from Anthropic and OpenAI - Opus 5, GPT 6.1 Sol, Luna and Astra - when post-trained on specific agentic tasks. Those tasks include document intelligence, financial accounting, scientific experimentation, medical insurance claims and security attack resilience.
Nvidia named Skydance, EY, Ramp, Crowdstrike, Perplexity and Salesforce as companies it says are using Super VL. Kari Briski, Nvidia's vice president of Open Models and Software, wrote in a blog post that any organization wanting to run agents on Nvidia infrastructure now has an open, optimized model for its tier without having to choose between capability and deployability. She also said that unlike closed frontier models, organizations can post-train Super VL on their own data, deploy it on their own infrastructure and own what they build.
The smaller Bolt models target a different audience. Nvidia says they are meant to run locally on the new AI PCs, aimed at consumers, programmers and enterprise knowledge workers, that are being released into the market now through partners including Microsoft and Dell. 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 smallest model, Bolt E4B, is aimed at conversational agents.
Briski wrote that PC users can delegate everyday projects without sending those workloads to the cloud. The pitch lands as privacy-weary consumers weigh how deeply to let AI assistants into their lives and give them access to information about their families, health and finances.
The launch extends Nvidia's existing Nemotron family, which already covers multimodal artificial intelligence, reasoning, coding, math and enterprise automation. Nvidia is leaning into demand from enterprises that want to run their own open models in their own data centers, even as fast-growing proprietary AI labs - companies that are both its customers and its investments - look to other chip and AI system makers, including AWS's Trainium and Cerebras, and build their own chips.
Super VL and Bolt 74B-A9B are available now, and Nvidia says Bolt E4B will arrive later this month.
More AI news from TechManNews.
