Open From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and Bolt Models
Source
blogs.nvidia.com
Author
Kari Briski
Date
Why it matters
Three open Nemotron models now cover server to edge AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → deployments: Super VL (120B) and Bolt 74B-A9B are available now, and Bolt E4B (4GB) arrives Oct. 23. Teams can pick a tier for local or hosted agents.
Key takeaways · AI-distilled
Bolt 74B-A9B uses native 2-bit quantizationShrinking a model by storing its numbers less precisely — like rounding — so it runs faster and fits on smaller hardware, at a small quality cost.Full definition → to fit in 24GB so it can run on RTX or DGX Spark alongside other apps. NVIDIA says it ranks #1 among open models in its class on PinchBench and #2 on BrowseComp.
Bolt E4B targets physical and conversational agents such as local voice assistants and embedded interfaces, running on GeForce RTX, Jetson, and RTX/DGX Spark.
Bolt 74B-A9B is on Hugging Face now with support for llama.cpp, Ollama, LM Studio, Unsloth, and vLLM. Super VL is on Hugging Face, OpenRouter, ModelScope, and as an NVIDIA NIM refreshed every two weeks.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
quantization — Shrinking a model by storing its numbers less precisely — like rounding — so it runs faster and fits on smaller hardware, at a small quality cost.
