Microsoft Azure documentation
Hugging Face on Microsoft Foundry
Hugging Face on Microsoft Foundry
Hugging Face has partnered with Microsoft to bring open-source models from the Hugging Face Hub into Microsoft Foundry and Azure Machine Learning. The Hugging Face Hub is the home of over 1,700,000 public access open-source models, as well as datasets, spaces and much more. The integration with Microsoft Foundry and Azure Machine Learning enables you to deploy open-source models of your choice to secure and scalable inference infrastructure on Azure powered by Hugging Face and other open-source inference solutions such as Text Generation Inference (TGI), vLLM or SGLang for LLMs and VLMs, or Text Embeddings Inference (TEI) for embeddings, and many more. Now, the Microsoft Foundry and Azure Machine Learning model catalogs are the home for over 11,000 of the most popular and downloaded open-source models on the Hugging Face Hub, ensuring secure and verified weights, that can be deployed to managed online endpoints with ease. Once deployed, the managed online endpoint gives you secure REST API to score your model in real time.
The Microsoft Foundry and Azure Machine Learning model catalogs contain over 11,000 Hugging Face deployable models, ranging a wide-variety of models for different tasks such as image generation, Large Language Models (LLMs), Visual Language Models (VLMs), or embeddings, among many others; all of those powered by open-source inference solutions. Additionally, each of those models can be deployed in a wide variety of hardware available on Microsoft Azure, ranging NVIDIA GPUs to CPUs, so that each model comes with a default suggested hardware.
At Microsoft Build 2025, an expansion of the partnership between Hugging Face and Microsoft Azure was announced. Among the main takeaways, the expanded collaboration will not only cover Azure Machine Learning, but also Azure AI Foundry Hub allowing Microsoft Azure users to design, customize, and manage AI apps and agents at scale with open-source models from Hugging Face.
Later that year, at Microsoft Ignite 2025, Azure AI Foundry was renamed to Microsoft Foundry, along with the announcement of a new experience, for a unified Azure platform-as-a-service offering for enterprise AI operations, model builders, and application development.
Read more about Azure AI Foundry evolving into Microsoft Foundry and other announcements at Microsoft Ignite 2025 in the Microsoft Ignite Book of news.
Resources
- Azure Machine Learning - Deploy models from Hugging Face Hub to Azure Machine Learning online endpoints for real-time inference
- Azure Machine Learning - How to use Open Source foundation models curated by Azure Machine Learning
- Microsoft and Hugging Face expand collaboration
- Microsoft and Hugging Face expand collaboration to accelerate Open-Source AI Innovation on Microsoft Foundry