The Apple on-device Foundation Model is the language model at the core of Apple Intelligence, the AI system in Apple's operating systems. Apple describes it as a compact model of about 3 billion parameters (the learned numbers inside a neural network), optimized to run on Apple silicon. App developers reach it through the Foundation Models framework, a Swift API; in code the model is the SystemLanguageModel class.
The model runs on the device itself, built for low-latency inference with minimal resource use. It is meant for text tasks such as summarization, entity extraction, text understanding, refinement and short dialog, and Apple states that it is not a chatbot for general world knowledge. Guided generation makes the model fill a developer's own Swift data structures, and tool calling lets it call functions in the app. Apple updates the model in routine operating system updates, and it works only on devices and regions that support Apple Intelligence.
Apple describes the on-device inference as free of cost; publishing an app on the App Store requires the paid Apple Developer Program.