foundation-models-on-device

You can run on-device LLMs in iOS by integrating the Apple FoundationModels framework to start a language model session, enabling private text generation and offline inference without sending data to the cloud.

Get structured outputs from an on-device iOS language model by applying the @Generable pattern within the FoundationModels framework to parse data and stream snapshots directly on the device.

Yes, you can use tool calling with Apple FoundationModels to execute domain-specific actions entirely offline, enabling robust custom interactions within your private iOS applications.

Snapshot streaming in on-device language models is a mechanism for receiving iterative generable outputs directly on an iOS device, needed when building responsive private assistants without cloud dependencies.

Yes, FoundationModels supports offline inference and private session management, allowing iOS applications to execute on-device text generation while keeping all user data completely local.

Limitations of using on-device LLMs for iOS text generation include relying entirely on local device hardware capabilities and requiring robust session management within the FoundationModels framework to handle offline constraints.