Nvidia Launches 120B-Parameter Nemotron 3.5 Super VL and Bolt Models for Edge AI
Updated
Updated · TechCrunch · Oct 19
Nvidia Launches 120B-Parameter Nemotron 3.5 Super VL and Bolt Models for Edge AI
3 articles · Updated · TechCrunch · Oct 19
Summary
Nvidia on Thursday released three open-weight AI models for always-on agents, led by the 120-billion-parameter Nemotron 3.5 Super VL for enterprises and two smaller Bolt models for local edge use.
Super VL is pitched as matching the larger Nemotron 3 Ultra on instruction-following and tool use while running faster on agentic tasks, with Nvidia claiming wins over Anthropic and OpenAI models after post-training.
Bolt 74B-A9B is designed for AI PCs and DGX Spark workstations to handle multistep reasoning, tool calling and web browsing locally, while the smaller Bolt E4B targets conversational agents and arrives later this month.
Microsoft and Dell are already shipping RTX Spark-based AI PCs, part of Nvidia's push to let users run assistants on-device rather than in the cloud for privacy-sensitive work.
The launch deepens Nvidia's bet on open models that customers can fine-tune and deploy on their own infrastructure as major AI labs increasingly explore rival chips and in-house silicon.