Apple’s Foundation Model framework has brought the power of large language models to the one device that people keep close for the majority of their day - the iPhone. And it’s enabled developers to build intelligent features using a privacy focussed framework - at zero cost for developers. But it does so much more - from type-safe guided generation using the @Generable macro to native tool calling. And now, with the ability to tap into the power of cloud-hosted models like Apple’s own Private Cloud Compute and third party models like Claude and Gemini, developers can architect applications that make the best use of the capabilities of the most suitable model.

This talk provides a comprehensive blueprint for building intelligent features using the LanguageModelSession API.