Microsoft brings AI model building to macOS with new Foundry tool

On May 21, Microsoft launched Foundry, a new AI development tool for macOS that lets users build their own small language models locally. Built into the Windows AI Studio platform, Foundry works directly on Apple Silicon Macs, opening up access for developers outside the Windows ecosystem.

Foundry focuses on giving users more control over model building by letting them start with existing small models like Phi or Mistral. Developers can then customize and refine them using their own datasets. Microsoft has positioned this as an option for companies that want AI tools without sending data to external servers.

Also read: Microsoft open-sources Windows Subsystem for Linux after nearly a decade

By offering a local workflow, Foundry reduces reliance on cloud infrastructure, which may appeal to companies concerned about privacy or latency. It also means developers can work offline, a feature that could speed up testing and experimentation.

While Windows AI Studio remains the full development suite for Microsoft’s AI efforts, Foundry’s macOS support signals the company’s effort to meet developers where they work. This move also helps Microsoft compete with Apple’s growing push into on-device AI, which is expected to expand at WWDC in June.

Microsoft also held its developer conference, Build 2025, which featured several announcements, such as AI chatbots for websites. You can read more about them here –

• Microsoft Engineer Disrupts Build 2025 Keynote Over Israel Contracts
• Microsoft Build 2025: Top 5 announcements highlights from the developer conference
• Microsoft Build 2025: Everything You Need to Know About Microsoft’s AI chatbots for Websites