Microsoft has opened the doors for developers to build and run AI models directly on macOS. With the launch of Foundry Local, macOS users can now create, manage, and deploy high-performance, on-device AI applications without needing constant cloud connectivity.
Foundry Local is Microsoft’s new local AI runtime stack, now available in preview for both Windows and macOS. It brings Azure AI Foundry’s core tools and models to your machine, allowing you to run inference workloads across CPUs, GPUs, and NPUs, including full GPU acceleration on Apple silicon. This shift puts powerful AI capabilities in your hands, even when you’re offline.
Cross-Platform AI, Built for Performance
Foundry Local runs on ONNX Runtime, optimized through tight integration with chipmakers including Apple, Intel, AMD, Qualcomm, and NVIDIA. It chooses the best hardware path automatically or lets you set it manually. The result: fast and efficient model execution, whether you’re building for desktops, laptops, or edge devices.
You can run a range of edge-optimized models like DeepSeek R1, Phi-4, Qwen 2.5, and Mistral, or load custom ONNX models from Hugging Face. Using Foundry Local Management Service, these models download and run at runtime without manual setup. Developers can use the new Foundry CLI to control local tools and agents or connect through the SDKs and OpenAI-compatible APIs. The stack also supports intelligent local agents using the Model Context Protocol (MCP), giving your apps smarter automation, entirely on-device.
How to Get Started on macOS
Installing Foundry Local on macOS is straightforward. Open Terminal and run:
brew tap microsoft/foundrylocal
brew install foundrylocal
foundry model run phi-3.5-miniThat’s all it takes to start running models locally.
According to Microsoft’s official blog, more than 100 customers are already using Foundry Local in private preview. Early users like SoftwareOne, Avanade, and iLink Digital have praised its reliability and hybrid AI capabilities for enterprise workflows.
Foundry Local turns your Mac into a capable AI machine. You can build fast, private, production-ready apps with full control over performance and deployment, without relying on the cloud.