Microsoft and Hugging Face expand collaboration

Source: HuggingFace Date: 2025-05-19 URL: https://huggingface.co/blog/azure-ai-foundry

Summary

Integration announcement: Microsoft and HuggingFace expand their partnership — 10,000+ curated HF models now available in Azure AI Foundry’s Model Catalog with one-click deployment. Security requirements for inclusion: ProtectAI Guardian scan, JFrog scanner, safetensors format, no remote code execution. Roadmap includes day-0 model releases, daily trending model updates, and expansion to video, 3D, time series, and protein modalities. Context: HF Hub now has ~2M models (up from 200k two years prior).

Implications

HF as open-source ML hub. 10,000 models in Azure AI Foundry makes HF the upstream supplier of AI models to Microsoft’s enterprise cloud — the most significant enterprise distribution integration to date. Day-0 release commitments mean new model releases will land in Azure simultaneously with HF Hub, removing the lag that previously forced enterprise teams to wait for Microsoft’s evaluation cycle.

Open-weights ecosystem health. The security requirements (Guardian scan, JFrog, safetensors, no remote code) establish a meaningful quality/safety bar for enterprise inclusion. Models that meet this bar gain access to Microsoft’s enterprise sales channel — a strong incentive for labs to ensure their releases are compliant from day one. Watch whether this creates a de facto “Azure-ready” certification in the community.