Apple plans to allow third-party developers to build applications using its large language models, which will be announced at the WWDC conference.
Apple Intelligence Foundation Models, currently used for various tasks like rewriting text, proofreading, generating images, and summarizing notifications, will be made available to developers through a software development kit and related frameworks.
Apple may differentiate its models by making them easier to use, potentially integrating them with its Core ML framework and offering guardrails to filter inaccurate AI output.
While Apple will initially offer on-device algorithms to developers, it's unclear if the more capable cloud-hosted large language models will also be made available. These large language models have shown promising performance and the ability to learn new skills using adapter layers.