NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-GGUF: 80GB+ VRAM

NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-GGUF: 80GB+ VRAM

At a Glance

Item Value
Repository lmstudio-community/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-GGUF
Published 2026-10-15
Formats GGUF
Source type Primary source (the publisher itself)

Values determined by this site’s code when the information was collected. Dates are JST.

Overview

The LM Studio community has released lmstudio-community/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-GGUF, a GGUF format conversion of the model “NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16" published by NVIDIA.

This repository is provided as part of the “LM Studio Community models highlights program," which features notable models within the community. It is a model supporting multimodal processing and conversational tasks, quantized utilizing llama.cpp, which is widely used in local inference engines. Additionally, it is noted that the NVFP4 version of the GGUF files was converted by the LM Studio team based on the “NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4" checkpoint released by NVIDIA.

Specifications

The model card description does not contain independent statements regarding specific specification numbers such as the breakdown of parameter counts, architecture details, or context length.

The quantized models provided in this repository were created by the LM Studio team utilizing the work of Georgi Gerganov and the llama.cpp development team. Furthermore, the disclaimer in the model card explicitly states that LM Studio is not the creator or owner of this model, but rather a third-party provided community model, and makes no guarantees regarding the accuracy, completeness, or safety of the output content.

Strengths and Use Cases

As indicated by the tags and model name assigned to the repository, this model, lmstudio-community/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-GGUF, has its primary strengths in multimodal processing and conversational tasks.

From the “VL" notation included in the model name and the tag structure, it is intended for use as a Vision-Language model that integrates and understands visual information such as images in addition to text. Through a conversational format with users, it is expected to be utilized in tasks such as question answering, explanations, and interactive information extraction based on visual input contexts.

On the other hand, the main text of the model card does not mention evaluation results or detailed use cases specialized for specific functions such as programming, mathematics, or agent tasks. Therefore, it cannot be confirmed from the documentation whether it specializes in specific expert fields, and it is appropriate to view it as a general-purpose multimodal conversational model capable of handling a wide range of vision-language dialogue tasks.

This model is a derivative model that converts NVIDIA’s original model “NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16" and its NVFP4 checkpoint into the GGUF format for easier execution in local environments. It is suited for easily trying out multimodal large conversational models locally in environments compatible with the llama.cpp ecosystem.

Hardware Requirements

Estimated requirements (calculated by Local Model Watch)

Your VRAM Quantization File size Est. memory needed
80GB class (A100 / H100) Q4_K_M 65.3GB 78.3GB

Memory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model’s authors. Compare with other models in our VRAM quick reference. What the quantization names mean: glossary.

Our Own Measurements

Measurements We Did Not Take

This model does not state a license, so we cannot confirm permission to use it. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).

How to Get It

This model is published in the Hugging Face repository lmstudio-community/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-GGUF.

The distribution format is GGUF. The LM Studio team performed the quantization work using llama.cpp, and in addition to the converted version from the BF16 original model, NVFP4 GGUF files converted based on the “NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4" checkpoint released by NVIDIA are provided.

Since this repository is not gated, anyone can directly download and obtain the model files without needing prior terms of service agreement procedures or approval applications on Hugging Face.

As for supported engines, it can be used with inference software that supports loading GGUF format files, including llama.cpp, which serves as the quantization conversion base, and LM Studio. When obtaining the model, you can download it using tools such as the Hugging Face CLI.

huggingface-cli download lmstudio-community/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-GGUF

As stated in the model card’s Disclaimers, LM Studio is not the creator or owner of this model, but rather in a position to highlight and provide models created by third parties as part of a community program. It should be noted that no guarantees are made regarding the accuracy, reliability, completeness, or safety of the content generated by the model, and users assume full responsibility for the results associated with downloading and using the model.

Related Articles

What to Read Next

Sources