At WWDC 2025, Apple made a groundbreaking announcement, unveiling the Foundation Models framework that gives developers direct access to Apple Intelligence for the first time. With this new framework, developers can integrate AI features like text generation, summarization, and suggestions directly within their apps, without having to rely on cloud services. 
Everything runs on-device, meaning user data remains private and no information needs to be sent to the cloud.
Apple claims developers can get started with just three lines of Swift code. This local, on-device approach provides fast performance and ensures that apps continue to work even when offline. Additionally, this setup maintains strict privacy standards while being free to use, with no associated costs for developers. Supported devices include the iPhone 15 Pro, iPhone 16 series, and any iPad or Mac with Apple Silicon.
In addition to text generation, the framework includes features like guided generation and tool calling, designed to help developers create smarter, more responsive apps. The Day One journaling app, for example, is already incorporating Apple Intelligence to improve user experience while maintaining privacy.
The framework is tightly integrated with Swift, allowing developers familiar with SwiftUI and Xcode 26 to adopt these features seamlessly without needing major changes to their apps. It’s a familiar toolset that opens up exciting new possibilities for app behaviors.
However, there are a few limitations. The models currently support a limited set of languages, though more will be added later this year. The framework also requires fairly recent hardware, meaning not every user will have immediate access to these features. Despite this, the move marks a significant shift in Apple’s developer strategy, as it opens up Apple Intelligence to third-party developers, who can now innovate with new AI-driven ideas.
While Apple’s local approach may not yet offer the same flexibility as cloud-based models from companies like OpenAI or Google, its focus on privacy and performance is likely to appeal to developers looking for more control over their AI applications.