Open From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and Bolt Models
The release includes open weights, training datasets, and recipes for a 120-billion-parameter multimodal model alongside two quantized edge variants.
- Super VL is a 120-billion-parameter vision-language model trained on 26 trillion tokens for multimodal agent workflows.
- Bolt 74B-A9B uses native 2-bit quantization to fit in 24GB of memory for local execution on RTX and DGX systems.
- Bolt E4B is a 4GB conversational agent model scheduled for release on October 23.
- The models and agent training datasets are available on Hugging Face, OpenRouter, ModelScope, and as NVIDIA NIM microservices.
Enterprises and developers can run and customize open agentic vision-language models across diverse hardware footprints, from 4GB edge devices to large data centers.
