Apple adds on-device generative AI SDK to iOS and macOS platforms

AI · 4 min read

Apple’s new Generative Engine SDK brings a standardized on-device inference layer to iOS and macOS, built on top of Core ML and Apple Neural Engine acceleration. The SDK exposes a consistent prompt-and-response API that supports text completion, image generation from prompts plus reference images, and short-form audio synthesis optimized for low latency.

Privacy is a headline feature: models run locally by default and Apple emphasized that system telemetry is opt-in and privacy-preserving. The SDK also includes model signing and attestation so apps can verify model provenance and ensure integrity when loading third-party or enterprise models.

For developers, Apple provides sample models, a conversion toolkit for popular open formats, and a new Xcode Device Profiler that visualizes memory, compute and power consumption of generative workloads. Early partners in creative and productivity categories say the local inference path reduces network costs and enables richer offline experiences.