TAO Production Branch October 2025 (PB 25h2) offers a 9-month lifecycle for API stability, with monthly patches for high and critical software vulnerabilities. This release includes Government Ready images for regulated environments.
What is TAO?
TAO (Train Adapt Optimize) Toolkit is a Python-based AI toolkit built on PyTorch that provides transfer learning capabilities to adapt popular neural network architectures and backbones to your data. The toolkit enables you to train, fine-tune, prune, quantize and export highly optimized and accurate AI models for edge deployment. TAO Toolkit can distill knowledge from large foundation models like CRADIOv2 and ConvNext-L's to smaller compute-friendly models for the edge.
The purpose-built pre-trained models accelerate the AI training process and reduce costs associated with large-scale data collection, labeling, and training models from scratch. Transfer learning with pre-trained models can be used for AI applications in smart cities, retail, healthcare, industrial inspection and more. TAO supports training for Computer Vision (CV) and 3D Point cloud modalities.
TAO packages a collection of containers, Python wheels, models and Helm charts. AI training tasks run either on PyTorch depending upon the entrypoint for the model. For deployment, TAO models can be deployed to DeepStream for video analytics applications, or Triton for inference serving use cases.
The container components are ready for commercial/non-commercial use.
What is TAO Production Branch October 2025?
The TAO Production Branch, exclusively available with NVIDIA AI Enterprise, is a 9-month supported, API-stable branch that includes monthly fixes for high and critical software vulnerabilities. This branch provides a stable and secure environment for building your mission-critical AI applications. The TAO production branch releases every six months with a three-month overlap in between two releases.
Getting Started with TAO Production Branch
Before you start, ensure that your environment is set up by following one of the deployment guides available in the NVIDIA AI Enterprise Documentation.
For an overview of the features included in the TAO Production Branch October 2025, please refer to the Release Notes for TAO 6.0.0.
Additionally, if you're looking for information on Docker containers and guidance on running a container, review the Containers For Deep Learning Frameworks User Guide.
Government Ready: STIG/FIPS Hardening
This ensures the highest level of security for regulated environments, the x86 container image for this branch is:
- STIG Ubuntu 24.04 hardened
- Supports FIPS 140-2 / 3 validated crypto / uses libraries that support FIPS crypto
To use this specific hardened image, navigate to the repository's Tags tab and look for the purple label indicating Gov ready displayed alongside the tag.
Learn more about NVIDIA's hardened image in the AI Software for Regulated Environments White Paper.
Governing Terms
The software and materials are governed by the NVIDIA Software License Agreement and the Product-Specific Terms for NVIDIA AI Products.
You are responsible for ensuring that your use of NVIDIA provided models complies with all applicable laws.
Deployment Geography:
Global
Release Date:
NGC [10/27/2025] via [https://catalog.ngc.nvidia.com/orgs/nvidia/containers/tao-pb25h2]
Program Classes:
TAO includes the following models and components:
| Model Name & Link | Use Case | How to Pull the Model |
|---|---|---|
| CRADIOv2 | Multi-teacher distilled foundation model generating rich visual embeddings | Manual |
| ConvNextv2 | FC-MAE trained foundation model generating rich visual embeddings for CNNs | Manual |
| TrafficCamNet Transformer Lite | Object detection network for detecting 4 class objects in traffic scenes | Manual |
| DeiT-base | Pretrained feature network for Projected GAN | Automatic |
| EfficientNet | Pretrained feature network for Projected GAN | Automatic |
| DeiT-small | Classifier guidance to inject class information | Automatic |
| InceptionNet | Calculating FID scores/metrics | Manual |
| EfficientNet-embeddings | Prevent class embeddings collapse for optimization of the embeddings | Manual |
TAO Containers
All containers needed to run TAO can be pulled from this location. See the list below for all available containers in this registry.
| TAO Container Type | container_name:tag | What's it used for? |
|---|---|---|
| TAO PyTorch container | nvcr.io/nvidia/tao-pb25h2:6.0.1-pytorch | Finetuning workflows in PyTorch |
| TAO Deploy container | nvcr.io/nvidia/tao-pb25h2:6.0.1-deploy | TensorRT inference workflows in Deploy |
| TAO Data Services container | nvcr.io/nvidia/tao-pb25h2:6.0.1-data-services | Dataset augmentation, auto labelling and analysis workflows |
Deployment Details:
To get started with TAO, download the tutorial notebook using the NGC CLI and run the following commands.
Note: The instructions to install the NGC CLI are detailed here.
git clone https://github.com/NVIDIA/tao_tutorials.git
git checkout release/6.0.0
You can kick-start the notebook by running
python3 -m pip install nvidia-tao==6.0.1
python3 -m pip install jupyter
jupyter notebook --ip 0.0.0.0 --port <port_number> --allow-root
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.
DeiT-base, EfficientNet, and DeiT-small are pulled automatically through the timm API in the training code. CRADIOv2, InceptionNet and EfficientNet-embeddings need to be pulled manually. The tutorial notebooks outline the instructions and steps to pull these models.
Reference(s):
- Improve accuracy and robustness of vision ai models with vision transformers and NVIDIA TAO
- Train like a 'pro' without being an AI expert using TAO AutoML
- Create Custom AI models using NVIDIA TAO Toolkit with Azure Machine Learning
- Developing and Deploying AI-powered Robots with NVIDIA Isaac Sim and NVIDIA TAO
- More information about TAO Toolkit and pre-trained models can be found at the NVIDIA Developer Zone
- TAO documentation
- TAO getting Started guide and release notes
- TAO Toolkit Developer Forums
- DeepStream SDK
- Riva for ConvAI use case
Container Version(s):
TAO Toolkit - Complete package including 3 containers (PyTorch, Deploy, Data Services), tutorial notebooks, NVIDIA TAO launcher wheel, TAO Helm Chart, CRADIOv2 foundation models, ConvNext series foundation models, and TrafficCamNet Transformer Lite model.
Compatible Infrastructure Software Versions
For the optimized performance, it is highly recommended to deploy the supported NVIDIA AI Enterprise Infrastructure software in conjunction with your AI software. Production Branch - October 2025 (25h2) is compatible with NVIDIA AI Enterprise Infrastructure 7.
Security Vulnerabilities in Open Source Packages
Please review the Security Scanning tab on NGC to view the latest security scan results. For certain open-source vulnerabilities listed in the scan results, NVIDIA provides a response in the form of a Vulnerability Exploitability eXchange (VEX) document. The VEX information can be reviewed and downloaded from the Security Scanning tab.
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 developer team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.
Please report security vulnerabilities or NVIDIA AI Concerns here.
Get Help
Enterprise Support
Get access to knowledge base articles and support cases or submit a ticket.
NVIDIA AI Enterprise Documentation
Visit the NVIDIA AI Enterprise Documentation Hub for release documentation, deployment guides and more.
NVIDIA Licensing Portal
Go to the NVIDIA Licensing Portal to manage your software licenses.