Apple highlights third-party apps using its on-device Apple Intelligence models

These are some of the iPhone, iPad, and Mac apps that can use Apple's on-device Apple Intelligence models to create new experiences.

Apple announced the new Foundation Models framework at the WWDC 2025 developer event earlier this year, which enables third-party developers to access the company's on-device foundation models for free and improve their apps.

These foundation models power Apple's Intelligence on iPhones, and developers can use them to create AI-powered app experiences that also work offline. Developers can generate personalized quizzes for students and create insightful workout summaries, among various use cases.

Apple released the iOS 26, iPadOS 26, and macOS 26 software updates earlier this month, bringing the framework to compatible iPhones, iPads, and Macs. The company has highlighted several apps that already use its Foundation Models framework.

An app called Detail: AI Video Editor uses the framework to convert a draft into a ready-to-record teleprompter script. The app can generate a title, description, hashtags, and messages when a video is ready to share.

SmartGym utilizes it to generate summaries, including monthly progress overviews, routine breakdowns, and individual exercise performance reports. The journaling app called Stoic generates hyperpersonal journaling prompts from recent entries, suggests contextual journaling prompts, and provides starting phrases for a new entry.

A biology app called CellWalk utilizes the framework to generate a conversational explanation of a term, grounding the responses in the app's scientific information. The to-do app Stuff utilizes Apple's foundation models to power Listen Mode, which creates entries using voice commands, and Scan Mode, which captures handwritten tasks.

Apple previously explained that the framework has native support for Swift, and app developers can use Apple Intelligence models "with as few as three lines of code." It supports guided generation to ensure that models respond with a consistent format and structure that is useful to developers.

On the other hand, developers can grant access to their app's specific tools and functions so that the model can call back into the app when it needs more information for processing.