Apple Launches Python SDK for On-Device AI, Bringing Foundation Models to macOS 26 Developers
Summary
Apple launches a Python SDK for its Foundation Models framework, enabling developers to run on-device AI inference powered by Apple Intelligence directly from Python on macOS 26, with support for real-time text streaming, structured output, and batch inference — currently available in beta via GitHub.
Key Points
- Apple releases a Python SDK for its Foundation Models framework, enabling developers to run on-device AI inference powered by Apple Intelligence directly from Python code on macOS 26.0+.
- The SDK supports real-time text streaming, guided generation with structured output schemas, type-safe responses via Python decorators, and batch inference for evaluating Swift Foundation Models app features.
- Currently in beta and not yet open to contributions, the SDK requires macOS 26.0+, Xcode 26.0+, and Python 3.10+, and is available for installation from source via GitHub under an Apache-2.0 license.