I heard foundation models are a big deal. Can you give me an overview of their capabilities?

Foundation models are a significant advancement in AI, particularly in the context of Apple's ecosystem. At WWDC 2025, Apple introduced the Foundation Models framework, which provides access to an on-device large language model through a Swift API. This framework is available on macOS, iOS, iPadOS, and VisionOS, and it allows developers to enhance their apps with intelligent features while keeping data private and offline.

Key capabilities of the Foundation Models framework include:

1. Text Generation and Summarization: The framework can generate content, summarize text, and analyze user input, making it useful for applications like personalized search suggestions or generating itineraries in travel apps.

2. Structured Output: Using features like Generable and Guided Generation, developers can prompt the model to produce structured data, such as JSON or CSV, which is easier to integrate into app interfaces.

3. Tool Calling: This feature allows the model to autonomously execute code defined in your app, enabling it to perform actions like fetching live data or interacting with app-specific functions.

4. Stateful Sessions: The framework supports multi-turn interactions with stateful sessions, allowing for more complex and context-aware conversations.

5. Privacy and Efficiency: Since the model runs on-device, it ensures user data remains private and does not increase app size. The model is optimized for performance with techniques like quantization and speculative decoding.

For a more detailed overview, you can refer to the session Meet the Foundation Models framework (00:00:00) and Deep dive into the Foundation Models framework (00:00:00).