WWDC26: Platforms State of the Union | Apple

Apple Event.

Welcome to the 2026 Platforms State of the Union. This is one of our favorite moments of the year where we get to share what's new with the technologies, frameworks, and tools you use every day to build incredible apps and games. We love connecting with you, hearing about your passions and challenges, and how we can better support your work. Your feedback shaped the new design with Liquid Glass and Apple Intelligence — huge themes in the 26 releases and key again this year. Design and intelligence enhance what's special about your apps: the care, craft, unique interfaces, and rich experiences shaped by your domain expertise, combined with enhanced intelligence capabilities to build features that weren't previously possible. We'll cover three key areas. First, Apple Intelligence — new ways to bring generative intelligence into your apps and new integrations with system intelligence to bring users back to your apps. Second, platform improvements — design refinements, flexible UI layout, updates to Swift and SwiftUI, and enhancements that make your apps faster, more adaptive, and easier to build. And finally, developer productivity — agentic coding taken further, alongside improvements that make Xcode faster and more personal. Let's get started with Apple Intelligence. At the heart of Apple Intelligence are Apple Foundation Models, created with Google leveraging their Gemini family of models, adapted to run on device and on Private Cloud Compute. Through the Foundation Models framework, your apps can use them too. This year, the framework expands to include image input and server model support — the API can integrate with the cloud model provider of your choice. Developers with fewer than 2 million first-time App Store downloads can use Apple Foundation Models on Private Cloud Compute with no cloud API cost — frontier-level intelligence with unparalleled privacy protections. The App Intents framework connects your app to Apple Intelligence through the Spotlight semantic index, the app toolbox, and the system orchestrator. Together, in-app and system-wide intelligence unlock experiences neither could deliver alone. Here's Richard and Mary Beth.

The Foundation Models framework is a native Swift API giving you direct access to the same on-device model that powers Apple Intelligence. Many of you have already adopted it — shopping apps like Wayfair, educational apps like CellWalk, local sports apps like CricHeroes — all running on device with no infrastructure costs or privacy trade-offs. It's amazing to see how you've pushed the limits of what an on-device model can do. We've got exciting updates. Let's start with a preview of the intelligence-powered features you'll be able to build today. Mary Beth, over to you.

Mary Beth presents the Origami sample app — a place to unwind with Japanese paper craft. The app features a gallery of origami projects tailored to each person's interests and materials using foundation models, with step-by-step feedback and a built-in chat for planning meetups. She demos combining a friend's paper with a photo of her dog to generate a fun project — the model analyzes inspiration pictures, translates Japanese text, and brainstorms folding options. The Foundation Models framework expands this year with multimodal prompts supporting text and images, plus Vision framework integration for OCR and barcode reading on device. Server models can now be called for complex workflows, supporting Claude, Gemini, and any provider conforming to the Language Model protocol. Developers getting started with AI can use Apple Foundation Models on Private Cloud Compute with no cloud API cost, while iCloud+ subscribers get expanded access. A new open source Swift package provides pre-built tools with skills and utilities for context management. Dynamic Profiles — new declarative APIs — enable truly adaptive AI experiences with less code, letting you orchestrate skills and sub-agents, swap tools, and update instructions on the fly. In code, a LanguageModelSession uses profile switching: a brainstorming profile using Private Cloud Compute with high creativity, a tutorial generation profile with deep reasoning, and an on-device profile for explaining origami jargon. Profiles share a continuous transcript for more contextual intelligence with less prompting, and their bodies recompute on every model turn. Dynamic Profiles are adaptable building blocks for AI agents and skills. New developer tools include the Evaluations framework for testing prompts, an upgraded Foundation Models instrument for debugging, a command-line tool for terminal prompting, a Python SDK, tool calling with images, and a RAG tool powered by Core Spotlight. The Foundation Models framework will be open source later this summer, enabling the same Swift APIs to run on servers for complete end-to-end AI workflows.

You've seen how the Foundation Models framework connects to third-party models, Private Cloud Compute, and the on-device model. You have the flexibility to get the right model for the job. And when you want to bring a specific model into your app and run it on device, there's Core AI — a brand new framework built into the platform. It delivers uncompromising performance through a modern memory-safe Swift API, with fine-grained resource management, model specialization, and custom GPU kernels. Python-based tools let you convert and optimize PyTorch models for the Core AI runtime. The framework is backed by a new developer toolchain with ahead-of-time compilation, dedicated Core AI instruments, and a visual debugger that traces tensor values back to your original Python source code. It scales with your available compute — from a compact vision model on iPhone for real-time camera queries to a multi-billion parameter LLM on Mac for agentic assistants handling complex multi-step workflows. Everything runs on-device with zero server dependencies and zero token costs. Core AI is optimized for Apple Silicon and empowers Apple Intelligence experiences across the system, including Siri. And this year, Apple Intelligence offers even more opportunities for developers. Over to Lori.

Apple Intelligence draws on personal context from across apps, understands what's on screen, and can take actions. You can integrate with these capabilities through the App Intents framework. Entity schemas describe your app's content — contributing to the Spotlight semantic index makes information discoverable with attribution back to your app. Intent schemas describe actions Siri can perform naturally, covering categories like task management, photo editing, and communication. System-defined schemas benefit from future Siri language understanding updates automatically. Lori demos the Origami app with Siri. She asks who's coming to origami night — Siri responds: 'Based on your messages in Origami, it looks like Kevin, Mary Beth, Rachel, and Richard are discussing origami night.' She asks what Richard is bringing and Siri replies: 'Richard mentioned he is thinking of bringing pizza.' She then sends a text through Siri: 'Can you make one of the pizzas vegetarian?' Siri confirms and sends it. Using the new View Annotations API, she associates message views with entities so users can reference on-screen content conversationally — saying 'this photo' or 'the second message' — and Siri takes action. She sends a photo to Kevin with a message about paper practice, and Siri handles it. By combining personal context, common app actions, and on-screen awareness, your app becomes part of the intelligent fabric of the system — accessible through natural language, discoverable through semantic search, and integrated into daily workflows. Now, back to Josh.

This is our vision for an intelligent platform — rich native experiences and intelligent natural language interfaces working together. You can enhance your app through Siri, build powerful AI features with the Foundation Models framework, and run your own models on-device with Core AI. If you're using a custom model in your app, Core AI is the right technology — efficient across all devices and built into the platform so your apps always benefit from the latest fixes. For enthusiasts experimenting with, training, or fine-tuning generative models, or running local inference servers, MLX makes it easy to explore cutting-edge innovations. It now supports Metal 4, GPU Neural Accelerators, and multi-Mac training across Thunderbolt with RDMA — all open source and faster than ever. The breadth of Apple Intelligence powered by Apple Silicon makes our platforms the best place to build intelligence-enhanced apps and games. Now let's look deeper at what makes all of this possible — the systems your apps depend on. We took a close look at performance and quality, and you'll see platform improvements across the board. When you rebuild with the new SDK, apps launch faster and feel more responsive. You'll see refinements across frameworks, media, search, accessibility, Swift, SwiftUI, and especially design. Last year, Liquid Glass brought a unified design language built for experiences that move across devices, making apps more expressive and delightful while staying instantly familiar.

Cindy details Liquid Glass design updates for the 27 releases. Readability is improved with better content diffusion behind glass elements, while darkened edges and brighter specular highlights add depth and separation. A new Settings slider lets users personalize from ultra clear to fully tinted. These improvements apply automatically to apps already using Liquid Glass with no recompilation needed. The design adapts to accessibility settings like reduced transparency and increased contrast, and macOS 27 now supports the 'show borders' environment value. Sidebars expand to edges on Mac and iPad for clearer structure while still refracting content. Sidebar icons regain accent colors for more personality and window identification. Every macOS window gets a tighter corner radius for consistency. Uniform toolbars improve contrast when content scrolls under floating bars. Icons and menus can be shown intentionally for key actions on macOS and iPadOS. Icon rendering is updated with sharper, more defined Liquid Glass, including new refraction features. Icon Composer now supports multi-layer Liquid Glass design with annotation features and interactive preview across releases. For app adaptability, iOS apps can now be resized in iPhone Mirroring and on iPad. Apps are automatically opted in when rebuilt with the latest SDK. SwiftUI apps with scene lifecycle are already well-positioned. A resizable simulator, Previews, and a coding agent skill help test and fix resizability issues. Instead of designing for specific devices, you're designing for a dynamic range of sizes and aspect ratios.

SwiftUI is the best way to build apps for any Apple device. As new capabilities like Liquid Glass are added, SwiftUI apps get them easily. New apps like The Goat, a game development environment bringing the open source Godot engine to Apple devices, are built with SwiftUI. Apps like Notion are migrating their UI to SwiftUI for performance and UI consistency. New interactions: Reorderable containers make drag-to-reorder trivial — just add .reorderable() to ForEach and .reorderContainer() to the parent, and SwiftUI handles lift and drop animations across grids, stacks, and any container. Swipe actions now work inside any container. Text selection gains full-fidelity selection on iOS and custom text renderers, text vibrancy, and vertical text on macOS. Speed improvements: SwiftUI, AppKit, and UIKit now share a common foundation across controls, benefiting from shared improvements. Nested stack layouts resize up to twice as fast by short-circuiting unnecessary measurements. State objects are now lazy — initialized only when first loaded — and AsyncImage automatically caches content using standard HTTP caching. New capabilities: Finer toolbar control with visibilityPriority for important items, overflow menu container for less prominent actions, and topBarPinnedTrailing for trailing-edge anchoring. Prominent tab role pins tabs to the trailing edge. A new document infrastructure provides first-class URL access for customizable disk read/write with observation, Swift concurrency, and more. The Spatial Preview framework lets Mac apps extend into 3D space around Apple Vision Pro users, enabling real-time preview, editing, and sharing of 3D models.

Swift is designed to be the language you reach for at every layer of the stack — from mobile apps to internet-scale services to embedded firmware. Its performance, depth, and interoperability make it the natural successor to C and C++ for systems and server programming, while its approachability makes it ideal for apps and frameworks. Cross-platform tools on Linux, Windows, Android, and the web are available on Swift.org. Developers are extending Swift across their stacks: Flighty shares code between app and backend, GoodNotes uses WebAssembly for cross-platform reach, and Frameo uses Swift-Java interoperability. Apple has been migrating its own systems to Swift: Foundation, AppKit, UIKit, and WebKit (replacing core C++ components including the QUIC transport layer, to be open sourced). The TrueType font rendering engine was rewritten in Swift — not only memory safe but faster. Parts of the core operating system kernel are now written in Swift for the 27 releases. Swift 6.4 improves everyday workflows: suppress warnings in specific code sections, use 'anyAppleOS' for availability attributes, await inside defer blocks, and significantly improved type checking — code that previously hit 'unable to type check expression' errors will now often compile successfully or give more actionable errors. There's never been a better time to go full stack with Swift.

Those are the platform improvements in this year's releases. macOS Tahoe was the final release to support Intel Macs — the transition to Apple Silicon is now complete. You can ship Apple silicon-only binaries on the Mac App Store, reducing download size and focusing testing on a single architecture. With Liquid Glass refinements complete, we're removing support for the old design — once recompiled with Xcode 27, apps automatically use Liquid Glass. With improvements across the system, your apps and games will look and feel better than ever. Now let's turn to your productivity. Intelligence is transforming how you write code, add features, and build apps. Last year we brought AI coding assistance to Xcode, and so many of you embraced it. This space moves fast, so we've picked up the pace of releases. Earlier this year we brought coding agents to Xcode with tools for grabbing previews, searching documentation, and more, powered by the Model Context Protocol. Xcode connects to tools you already use, from Figma to GitHub, and includes built-in integration for agents from Anthropic, OpenAI, and now Google. Today, Xcode adds Agent Client Protocol support so you can bring any compatible agent. ACP and Gemini integration ship in an update to Xcode 26 available today. Here's Ken.

Xcode is where you build the best apps for Apple platforms. This year has two big stories: intelligence and the daily experience. Xcode is faster at loading projects, with fixed crashes, more reliable debug sessions, faster expression evaluation, and a console that handles intensive logging. Xcode 27 is 30% smaller, now Apple silicon-only, with agents, documentation, and components downloading in the background. Settings sync automatically via iCloud — importing from another Mac is seamless with Git config included. New project creation goes straight to the editor with no file name, bundle ID, or setup. The toolbar is fully customizable with the activity view tucked into the document title. New themes bring color throughout the entire app with gorgeous presets like Emerald, Neon Noir, and Coral Reef, each supporting light and dark mode, with different themes per project. Xcode Cloud builds are up to twice as fast, now supporting Apple Vision Pro and Metal on Apple silicon. Previews now show variations for any property in a grid — all states of your UI at a glance. The new Device Hub replaces Simulator, combining simulator and physical device management. It supports rotation, screenshots, dynamic resizing, device property controls, pinch-to-zoom, two-finger scrolling, and interaction with physical devices from the same interface — all the convenience of the simulator with the fidelity of real hardware.

What a great time to be a developer. Intelligence is transforming how you build apps. Xcode 27 takes the next big step in agentic coding, leveraging the best models and agents directly in Xcode. Agents are woven into every layer — understanding projects, searching documentation, building, testing, rendering previews with variants, interacting with simulators, localizing, and debugging. Every answer is grounded in Swift, SwiftUI, and Apple frameworks — making Xcode the best place to code with agents for Apple platforms. Kevin demos building a choose-your-own-adventure feature for the Origami app. He starts a conversation with the agent from the toolbar, describes the feature with specific requirements, and adds /plan for upfront design collaboration. The agent explores the project using Xcode tools, asks clarifying questions, and produces a plan in rendered markdown. Kevin refines it — adding the ability for his daughter to create custom settings and items — then kicks off implementation. Xcode shows all code and preview changes in real time. The result is a complete feature with gorgeous hero images that his daughter will love. For validation, agents can run tests, use playgrounds, check visual changes in Previews across modes and localizations, and interact with the app in the simulator — tapping, swiping, and typing to test combinations and generating screenshots. For localization, Xcode automatically adds French and agents translate strings in context. For crash fixing, agents pull top crashes from the Organizer, symbolicate logs, identify issues, reproduce crashes, make fixes, and validate them. Xcode 27 ships with Apple-built specialists as plugins: SwiftUI, accessibility, universal sizing, testing, and performance. Plugins can contain skills, MCP tools, and agents via Agent Client Protocol. Partners like Figma and GitHub offer one-click setup. Kevin demos implementing a Figma design in SwiftUI, refining variants, making it resizable, and posting a PR to GitHub — all through agentic coding. Back to you, Josh.

It's a huge year for Xcode and many of our other developer tools. The all-new Reality Composer Pro 3 is completely rebuilt for crafting production-ready 3D experiences using RealityKit, with character animations, realistic lighting, and live previews using Mac Virtual Display. Game Porting Toolkit gets a major update with AI skills for coding agents, dramatically cutting porting time. New Metal command line tools give agents direct control during development and debugging. Across the 2027 releases: the App Intents framework connects your app to Apple Intelligence; the Foundation Models framework and Core AI enable powerful generative intelligence features; platform improvements in design, Swift, and SwiftUI make apps faster, more flexible, and easier to build; and Xcode has expanded agentic coding support. Over 100 sessions dive deep into everything we've covered, available on the Apple Developer app, website, YouTube, and new this year on Bilibili. Sign up for Group Labs, online panels, and Q sessions with Apple engineers and designers. Connect on the Apple Developer Forums. You can meet with Apple worldwide through hands-on workshops, labs, and events. We're excited to announce our fifth Developer Center opening this fall in Berlin, joining Cupertino, Shanghai, Singapore, and Bengaluru. Whether this is your first WWDC or your 25th, thank you. Your work inspires and drives us. We use the apps and play the games you build. Our greatest hope is that everything we talked about today will enable your next great idea to come to life. We can't wait to see what you do next. Enjoy WWDC.

Apple Event.