foundation-models-on-device

You can integrate on-device LLMs into your iOS app by using the FoundationModels framework to generate text, summaries, and creative content locally without requiring network connectivity.

Using the FoundationModels framework for on-device generative AI requires iOS 26 or later to ensure local processing, privacy, and offline support.

You can get structured Swift data output from an on-device LLM by using the @Generable attribute, which enforces type safety and simplifies parsing for predefined Swift types.

Yes, you can stream partially generated structured responses using snapshot streaming, which enables real-time UI updates as the on-device LLM generates output.

You can allow an on-device LLM to invoke custom code by implementing tool calling, which extends the LLM's capabilities to execute domain-specific actions within your iOS application.

On-device LLM processing keeps user data completely private because all generative AI tasks are executed locally on the device, ensuring no sensitive data ever leaves the user's hardware.