NVIDIA Nemotron 3.5 Super VL 120B A12B NVFP4 needs about 91.1 GB of memory at its main file (a 78.7 GB file) with an 8K context window, which is more than a 48 GB GPU holds. Longer context needs more: at 32K tokens the KV cache alone adds about 47.8 GB.
NVIDIA Nemotron 3.5 Super VL 120B A12B NVFP4 – a vision‑language model with 120B parameters
NVIDIA Nemotron 3.5 Super VL 120B A12B NVFP4 is a vision‑language model quantized with NVFP4. It contains 120B total parameters, 12B active parameters, and supports up to 1M context tokens. The model is trained for reasoning, coding, tool use, document and video understanding.
Good at
- Vision‑language reasoning
- Tool‑based code generation
- Document and video understanding
Limits
- Requires large VRAM (≈91.1 GB at 8k context)
- Only one concurrent sequence supported on DGX Spark (GB10)
- No explicit vision support in llama.cpp
Best for
- Large‑scale vision‑language inference
- Complex reasoning tasks
- Video content analysis
Who it’s for. Engineers and researchers needing a high‑capacity vision‑language model for advanced reasoning and tool integration. The model is suitable for commercial or non‑commercial use under the openmdw‑1.1 license.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 78.7 GB | 91.1 GB | Too big | Get file → |
The available quant is unknown, with a file size of 78.69 GB. Estimated VRAM usage is 91.1 GB at 8k context. For systems with less VRAM, consider using a lower‑precision quant if available, but no other quants are listed.
What GPU do you need for NVIDIA Nemotron 3.5 Super VL 120B A12B NVFP4?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: no quant fits at an 8K context.
- 16 GB VRAM: no quant fits at an 8K context.
- 24 GB VRAM: no quant fits at an 8K context.
- 32 GB VRAM: no quant fits at an 8K context.
- 48 GB VRAM: no quant fits at an 8K context.
Specifications
Architecture from the GGUF header
Frequently asked questions about NVIDIA Nemotron 3.5 Super VL 120B A12B NVFP4
How much VRAM does NVIDIA Nemotron 3.5 Super VL 120B A12B NVFP4 need?
About 91.1 GB at its main file with an 8K context window. That is the 78.7 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run NVIDIA Nemotron 3.5 Super VL 120B A12B NVFP4 on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 91.1 GB, so you would need a larger GPU or CPU offloading.
What is the file size of NVIDIA Nemotron 3.5 Super VL 120B A12B NVFP4 GGUF?
It depends on the quant: file is 78.7 GB.
What context length does NVIDIA Nemotron 3.5 Super VL 120B A12B NVFP4 support?
The GGUF header lists a trained context length of 1,048,576 tokens (1024K). The KV cache grows with the context you use: about 11.9 GB at 8K and 47.8 GB at 32K.
Where can I download NVIDIA Nemotron 3.5 Super VL 120B A12B NVFP4 GGUF?
The files are on Hugging Face in the FedericoFB/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does NVIDIA Nemotron 3.5 Super VL 120B A12B NVFP4 use?
Hugging Face lists the license as openmdw-1.1. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run this model on 8 GB VRAM?
No. The model requires approximately 91.1 GB of VRAM at 8k context, far exceeding 8 GB.
Is the model open source?
Yes, it is released under the openmdw‑1.1 license and can be used commercially or non‑commercially.
Does the model support vision tasks?
Yes, it includes a C‑RADIO vision encoder and a vision‑language connector for vision‑language tasks.