NGC Catalog
CLASSIC
Welcome Guest
Models
Mistral-Nemotron-Vision-4B-Instruct

Mistral-Nemotron-Vision-4B-Instruct

For downloads and more information, please view on a desktop device.
Associated Products
Description
Mistral-Nemotron-Vision-4B-Instruct is a Vision Language Model capable of taking inputs of text, image(s) , video.
Publisher
NVIDIA
Latest Version
v1
Modified
July 24, 2025
Size
10.35 GB

Model Overview

Description:

Mistral-Nemotron-Vision-4B-Instruct is a versatile vision-language model designed for tasks such as document intelligence, function calling, and video understanding. It combines the Mistral-NeMo-Minitron-4B-Instruct language model with the RADIO vision encoder to deliver high accuracy across a wide range of RTX GPUs. This model is optimized for both commercial and non-commercial use.

This model was trained on commercial images and videos for all stages of training and supports single image and video inference.

License/Terms of Use

The use of this model is governed by the NVIDIA Community Model License, Additional Information: Apache License Version 2.0.

Deployment Geography

Global

Use Case

Customers: AI foundry enterprise customers

Use Cases: Image summarization. Text-image analysis, Optical Character Recognition, Interactive Q&A on images, Comparison and contrast of multiple images, function calling.

Release Date

[07/25/2025] [NGC]

Model Architecture:

Architecture Type: Transformer

Network Architecture

  • Vision Encoder: nvidia/CRadioV2-H
  • Language Encoder: nvidia/Mistral-NeMo-Minitron-4B-128K-Instruct

Input

Input Type(s): Video, Image(s), Text
Input Format(s): Video (.mp4), Image (Red, Green, Blue (RGB)), and Text (String)
Input Parameters: Video (3D), Image (2D), Text (1D)
Other Properties Related to Input:

  • The model has a maximum of 16384 input tokens
  • Minimum Resolution: 32 × 32 pixels
  • 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
  • Alpha Channel: Not supported (no transparency)

Output

Output Type(s): Text
Output Format(s): String
Output Parameters: 1D
Other Properties Related to Output: Maximum output tokens can be specified by the user.

Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated system. 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.

Prompt Format:

Single Turn

<s>System
{system prompt}</s>
<s>User
<image>
{prompt}</s>
<s>Assistant\n
<s>System
{system prompt}</s>
<s>User
{prompt}</s>
<s>Assistant\n

Software Integration:

Runtime(s): AI Inference Manager (NVAIM) Version 1.0.0

Supported Hardware Microarchitecture Compatibility: H100 SXM 80GB

Supported Operating System(s): Linux

Model version: V1.0

Training & Evaluation:

NV-Pretraining and NV-Nemotron-SFT were used for training and evaluation

Data Collection Method by dataset:

  • Hybrid: Automated, Human

Labeling Method by dataset:

  • Hybrid: Automated, Human

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).

Properties:

We used commercial Pretraining and SFT datasets including images, texts, and videos for training this model.

Evaluation Dataset:

Data Collection Method by dataset:

  • Hybrid: Human, Automatic/Sensors

Labeling Method by dataset:

  • Hybrid: Human, Automatic/Sensors

Properties:

A collection of different benchmarks, including academic VQA benchmarks.

Image Benchmarks

Benchmark Accuracy
MMMU_val with chatGPT as a judge 47.22
InfoVQA Val 73.5
OCRBenchV2 English 51.11
VideoMME(16 frames) 54.6
Text VQA 82.27
AI2D 82.28
ChartQA 86.28
OCRbench 832
RealWorldQA 73.46

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.

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.

Outputs generated by these models may contain political content or other potentially misleading information, issues with content security and safety, or unwanted bias that is independent of our oversight.