Siri AI at WWDC 2026 (simonwillison.net)

At WWDC 2026, Apple announced significant advancements to Siri AI, leveraging a custom Gemini-derived model that operates on their Private Cloud Compute (PCC) infrastructure. Notably, Siri will utilize vision large language models (LLMs) to extract relevant information directly from the user's screen, circumventing the need for custom app integration. This development marks a substantial leap forward, given the previous limitations of vision LLMs in 2024, and signals a commitment to enhancing user experience through advanced AI capabilities. Additionally, Apple introduced the Core AI library, which aims to empower developers to optimize their models for Apple hardware effectively. This library integrates with Meta's open-source PyTorch ecosystem through the Core AI PyTorch Extensions, facilitating the transition of existing PyTorch models to run efficiently on Apple devices. The incorporation of advanced security and privacy protocols while leveraging NVIDIA hardware within Google Cloud enhances Apple's infrastructure for handling complex reasoning tasks. These innovations present a promising shift in Apple's AI strategy, potentially redefining user interaction with Siri and broader AI applications on Apple platforms.