Matt Wolfe explains how to run open-source AI models on your own computer using Microsoft's Windows AI Foundry. You can browse models from Azure AI and Hugging Face, cache them locally, and switch from cloud endpoints to local models with minimal code changes. Large models like GPT OSS 20B work if your hardware supports them.
Open source and open weight models are everywhere. You can run them completely offline and your data won't get stored or shared with anyone else. But how do you actually run one of them on your own computer? That's where Windows AI Foundry can help. It's a new platform from Microsoft that helps developers run AI models directly on their devices, not just in the cloud. Here's how it works with Foundry Local, you can browse and choose open source AI models from places like Azure AI and Hugging Face and run them directly on your device. And if you're already using Azure AI Foundry to run models in the cloud, you can easily pull the open source model from the cloud and run it off of a local cache using Foundry Local SDK without having to rewrite a bunch of code. Even large models like OpenAI's GPT OSS 20B are able to run right on your computer if you meet certain power requirements. The platform also includes tools like a command line interface to help you find models optimized for your hardware, and a software development kit that makes it easy to point your app to a local model instead of a cloud endpoint. Whether you're building something new or enhancing an existing app, Windows AI Foundry gives you the tools to use local AI models in a fast, secure and flexible way.
AI News: Dots, GPT-6.1 Sol, Sonnet 5.5, Gemini 4, and everything you need to know
