Apple opens on-device generative AI APIs to third-party developers

AI · 5 min read

Apple has opened a new set of on-device generative AI APIs in the latest developer platform release, enabling developers to run optimized models for text completion, image edit and multimodal tasks inside apps without round-tripping user data to cloud services.

The APIs expose model management, quantized runtimes that target the Neural Engine and per-request privacy controls that let apps declare inference constraints. Apple emphasized developer tooling: Xcode now includes a model profiler that visualizes memory, latency and neural pipeline stages for device-level optimization.

For UX designers, the big change is predictable, private AI across apps — designers can prototype intelligent completion, contextual image edits, and real-time summarization while relying on managed on-device constraints. Apple also published human-interface guidelines for generative interactions, covering consent prompts, provenance metadata and reversible edits.