What does new Foundation model in iOS 27 for developers? What can I do?

In the WWDC context, the Foundation Models framework is Apple’s Swift API for using Apple’s on-device AI models in your app. It was introduced for building intelligent features with privacy, offline support, and no cloud API costs.

What developers can do with it

You can use it to build features like:

• Summarization
• Text extraction and analysis
• Classification
• Personalized search suggestions
• Generating content such as itineraries, quizzes, or replies
• Dialogue for characters in games
• Tool calling so the model can interact with your app
• Structured output with guided generation
• Streaming responses
• Multi-turn / stateful sessions

What’s new in the newer sessions

The newer WWDC 2026 sessions say the framework is expanding to include:

• Image input / image understanding
• Support for server models alongside the on-device model
• A model abstraction layer so you can swap models more easily
• System tools like Vision and Spotlight
• Dynamic Profiles for agentic app experiences
• Evaluations for measuring quality and testing AI behavior
• Open source release and utilities package
• FM CLI and a Python SDK for terminal and scripting workflows

In practical terms, what can you do?

You can now:

• Prompt Apple’s on-device model from your app
• Switch to a server model for more complex tasks
• Pass images into prompts
• Use AI from the terminal with fm
• Use Python for automation and evaluation pipelines
• Build more agentic workflows that combine models and tools
• Measure and debug behavior with Evaluations and Instruments

If you want, I can also give you a simple “what can I build with it?” list or a difference between on-device model vs server model summary.