Apple's WWDC 2026 expanded on-device AI capabilities significantly. Foundation Models now supports image input alongside text, enabling multimodal use cases without shipping a custom vision model. A second, more capable on-device model was announced for higher-end devices, supporting text, image understanding, and speech. Private Cloud Compute provides a fallback path when local capacity is insufficient. Core AI is a new native framework for running custom local models on Apple silicon, distinct from Foundation Models. For React Native developers, Callstack is working on package updates to expose these new Apple APIs through the existing JavaScript AI stack, covering image understanding, structured extraction, on-device classification, and hybrid local/cloud flows.
