Llama-3.1-Nemotron-Nano-VL-8B-V1-FP4-QAD

Model Overview

Description

Llama-3.1-Nemotron-Nano-VL-8B-V1-FP4-QAD is the quantized version of the NVIDIA Llama Nemotron Nano VL model, which is an auto-regressive vision language model that uses an optimized transformer architecture. For more information, please check here. The NVIDIA Llama Nemotron Nano VL FP4 QAD model is quantized with TensorRT Model Optimizer.

This model was trained on commercial images using Quantization-aware Distillation (QAD).

This model is ready for commercial/non-commercial use.

License/Terms of Use

Governing Terms: Your use of the model is governed by the NVIDIA Open License Agreement. Additional Information: Llama 3.1 Community Model License. Built with Llama.

Deployment Geography:

Global

Use Case:

The intended users of this model are AI foundry enterprise customers, as well as researchers or developers. This model may be used for image summarization, text-image analysis, Optical Character Recognition, interactive Q&A on images, and Chain-of-Thought reasoning.

Release Date:

• Hugging Face [October 8th, 2025] via https://huggingface.co/nvidia/Llama-3.1-Nemotron-Nano-VL-8B-V1-FP4-QAD

Model Architecture:

Network Type: Transformer

Network Architecture:

Vision Encoder: C-RADIOv2-H

Language Encoder: Llama-3.1-8B-Instruct

Number of model parameters: 8 billion

Input

Input Type(s): Image, Text

Input Format(s): Image (Red, Green, Blue (RGB)), and Text (String)

Input Parameters: Image (Two-Dimensional - 2D), Text (One-Dimensional - 1D)

Other Properties Related to Input:

• Language Supported: English only
• Input + Output Token: 16K
• Maximum Resolution: Determined by a 12-tile layout constraint, with each tile being 512 × 512 pixels. This supports aspect ratios such as:
  • 4 × 3 layout: up to 2048 × 1536 pixels
  • 3 × 4 layout: up to 1536 × 2048 pixels
  • 2 × 6 layout: up to 1024 × 3072 pixels
  • 6 × 2 layout: up to 3072 × 1024 pixels
  • Other configurations allowed, provided total tiles ≤ 12
• Channel Count: 3 channels (RGB)
• Alpha Channel: Not supported (no transparency)

Output

Output Type(s): Text

Output Formats: String

Output Parameters: One-Dimensional (1D)

Other Properties Related to Output: Input + Output Token: 16K

Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA’s hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions.

Software Integration

Runtime Engine(s): vLLM
Supported Hardware Microarchitecture Compatibility: B100/B200
Supported Operating System(s): Linux

The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.

Model Versions:

Llama-3.1-Nemotron-Nano-VL-8B-V1-FP4-QAD

Quick Start

Install Dependencies

pip install transformers accelerate timm einops open-clip-torch

Usage

To serve this checkpoint with vLLM, you can start the docker vllm/vllm-openai:latest and run the sample command below:

python3 -m vllm.entrypoints.openai.api_server --model nvidia/Llama-3.1-Nemotron-Nano-VL-8B-V1-FP4-QAD --trust-remote-code --quantization modelopt_fp4

Training Dataset

Data Modality:

• Image
• Text

Image Training Data Size:

• 1 Million to 1 Billion Images

Text Training Data Size:

• Less than a Billion Tokens NV-Pretraining and NV-CosmosNemotron-SFT were used for training and evaluation

Data Collection Method by dataset:

• Hybrid: Human, Synthetic
  

Labeling Method by dataset:

• Hybrid: Human, Synthetic
  

Properties: The dataset collection (for training and evaluation) consists of a mix of internal and public datasets designed for training and evaluation across various tasks. It includes:
• Internal datasets built with public commercial images and internal labels, supporting tasks like conversation modeling and document analysis.
• Public datasets sourced from publicly available images and annotations, adapted for tasks such as image captioning and visual question answering.
• Synthetic datasets generated programmatically for specific tasks like tabular data understanding.
• Specialized datasets for safety alignment, function calling, and domain-specific tasks (e.g., science diagrams, financial question answering).

Evaluation Dataset:

NV-Pretraining and NV-CosmosNemotron-SFT were used for training and evaluation.

Data Collection Method by dataset:

• Hybrid: Human, Synthetic
  

Labeling Method by dataset:

• Hybrid: Human, Synthetic
  

Properties: Additionally, the dataset collection (for training and evaluation) consists of a mix of internal and public datasets designed for training and evaluation across various tasks. It includes:
• Internal datasets built with public commercial images and internal labels, supporting tasks like conversation modeling and document analysis.
• Public datasets sourced from publicly available images and annotations, adapted for tasks such as image captioning and visual question answering.
• Synthetic datasets generated programmatically for specific tasks like tabular data understanding.
• Specialized datasets for safety alignment, function calling, and domain-specific tasks (e.g., science diagrams, financial question answering).

Evaluation Benchmarks:

Benchmark	Score (FP4)	Score (BF16)
MMMU Val with chatGPT as a judge	47.9%	48.2%
AI2D	85.0%	85.0%
ChartQA	86.5%	86.3%
InfoVQA Val	77.6%	77.4%
OCRBench	836	839
OCRBenchV2 English	59.5%	60.1%
OCRBenchV2 Chinese	38.0%	37.9%
DocVQA val	91.5%	91.2%
VideoMME*	54.6%	54.7%

^*Calculated with 1 tile per image.

The evaluation for this checkpoint was done with FP4 simulated quantization on H100.

Inference

Engine: vLLM
Test Hardware:

• 1x NVIDIA B100/B200

Ethical Considerations:

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse. For more detailed information on ethical considerations for this model, please see the Model Card++ Explainability, Bias, Safety & Security, and Privacy Subcards. Please report security vulnerabilities or NVIDIA AI Concerns here. Please make sure you have proper rights and permissions for all input image and video content; if image or video includes people, personal health information, or intellectual property, the image or video generated will not blur or maintain proportions of image subjects included. Users are responsible for model inputs and outputs. Users are responsible for ensuring safe integration of this model, including implementing guardrails as well as other safety mechanisms, prior to deployment. Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns here.