At WWDC 2026 Apple introduced Core AI, a developer framework that provides Swift-native access to Apple’s on-device foundation models. The framework is designed so models run locally on users’ devices, eliminating cloud round-trips and allowing developers to add intelligence to iOS and macOS apps without external API calls.
What Core AI offers
- Tool calling in Swift: developers can define custom tools in Swift that the on-device model can invoke, enabling agent-like workflows to run entirely on the user’s device without a server.
- Structured generation: models can return typed, schema-conforming outputs rather than only free text, reducing the need for post-processing and making features more reliable.
- Full Apple Intelligence integration: Core AI integrates with Apple Intelligence so AI features can be coordinated across apps by the system.
Privacy and no cloud dependency
Apple emphasizes a privacy-first approach: models execute locally on the device, which means no API keys, no usage-based pricing, and no latency from network calls. According to Apple, user data remains on the device.
Gemini models and Google collaboration
Apple also announced that Apple Intelligence now runs on foundation models co-developed with Google under the Gemini name. These models provide multimodal capabilities including image generation, visual question-and-answer, and speech generation. Gemini-based models can run on-device or via Private Cloud Compute.
System orchestrator and OS integration
A system orchestrator will automatically coordinate AI features across applications so that intelligence functions work seamlessly within iOS and macOS.
Availability and documentation
Core AI ships with iOS 27, macOS 27, and the latest Xcode. Developer documentation is available at developer.apple.com/documentation/coreai.
Why it matters
Core AI changes how developers can add intelligence to apps by removing reliance on external APIs and cloud infrastructure. Local model execution reduces network dependence, simplifies cost and privacy considerations, lowers latency, and—together with the Google Gemini collaboration—expands the platform’s multimodal capabilities.
Summary
Core AI delivers Swift-native, on-device AI development for iOS 27 and macOS 27 with tool calling, structured outputs, and privacy-first local execution. Apple Intelligence now leverages Gemini models co-developed with Google and supports both on-device and Private Cloud Compute deployments.
