Hugging Face open‑source models to production on Microsoft Foundry | DEM320
Microsoft Foundry introduces Foundry Managed Compute, a managed PaaS for deploying open-source and custom models.
“We have this construct called foundry managed compute which is think of it as a managed platform as a service where you just pick the GPUs.”
At Microsoft Build, the Foundry team announced Foundry Managed Compute, a managed platform-as-a-service that lets developers deploy open-source and custom models by simply selecting GPUs while Foundry handles infrastructure, runtimes (vLLM, SGLang), and maintenance. Backed by a Hugging Face partnership giving access to 10,000+ models, it lowers the barrier to productionizing open-source models on the same endpoint, SDK, and billing as proprietary frontier models. This matters because it signals Microsoft's push to make the open-source model ecosystem a first-class production path alongside proprietary models.