Apple has unveiled its Foundation Models framework, granting developers access to its on-device AI models. This framework allows for offline AI processing, enabling features like on-device quiz generation from notes or natural language search in apps, all while prioritising user privacy and eliminating inference costs. The models, which power Apple Intelligence, are optimised for Apple silicon and include a compact, approximately 3-billion-parameter model.
The framework natively supports Swift, allowing developers to integrate the Apple Intelligence model with minimal code. It also incorporates built-in safety guardrails to mitigate harmful model input and output. Apple is providing resources such as Generative AI Human Interface Guidelines for Responsible AI principles to aid developers in incorporating AI safety into their apps.
This move marks a significant step in bringing AI capabilities directly to applications, enhancing user experience with intelligent features that are both powerful and privacy-focused. The framework is currently available for testing through the Apple Developer Program, with a public beta planned for next month.
