Core AI is designed to be the best way to bring and run AI models on device in your apps. It’s a complete set of technologies purpose-built for Apple Silicon, engineered for uncompromising performance, extensive customization, and seamless scaling across devices and model sizes, with zero server dependencies and zero token costs.
Build intelligent experiences
Create intelligent features and enable new experiences for your apps and games by leveraging powerful on-device machine learning. Learn how to build, use, train, and deploy AI and machine learning models for iPhone, iPad, Apple Vision Pro, Mac, and Apple Watch.
Get to know the technologies
Core AI
Foundation Models framework
The Foundation Models framework is a native Swift API that gives you direct access to Apple Foundation Models — on device and in Private Cloud Compute* — as well as any model provider with a Swift package conforming to the Language Model protocol.
From there, multimodal prompts and on-device Vision framework tools let your app reason about images alongside text, while Dynamic Profiles let you swap models, tools, and instructions within a continuous session so your app's intelligence can adapt in real time. Then use the Evaluations framework to ensure your AI features work reliably across dynamic conditions.
Vision
Build powerful image and video analysis with the latest in computer vision. Tap-to-segment lets you isolate objects within images, while OCR, barcode scanning, and your own custom tools can be passed directly to Apple Foundation Models, enabling LLM-powered visual understanding in your app. Vision is also available on watchOS, bringing image analysis capabilities across Apple platforms.
Speech
Take advantage of speech recognition and saliency features for a variety of languages. With SpeechAnalyzer, you can bring advanced, on-device transcription to your app.
Machine-learning-powered APIs
Bring intelligent on-device machine-learning-powered features, natural language analysis, translation, and sound classification, to your app with just a few lines of code.
Core ML
Core ML delivers fast performance for integrating traditional machine learning models into your apps and games — from tree ensembles to regression models and beyond. Convert models from popular training libraries using Core ML Tools, download ready-to-use models, and preview them directly in Xcode. If you're interested in bringing LLMs and other generative AI models into your app, check out Core AI.
Metal
Metal puts the advanced capabilities of Apple-designed GPUs at your fingertips to power the most advanced graphics workloads. Now you can tap into machine learning capabilities like MetalFX, run inference networks directly in your shaders, and implement the latest neural rendering techniques with Metal 4.
MLX framework
MLX is an open-source array framework that lets you experiment with, training, researching, and fine-tuning generative models on Apple Silicon. It supports Metal 4 and GPU Neural Accelerators for maximum performance, and can scale training across multiple Macs with RDMA over Thunderbolt, making it an incredible way to explore cutting-edge machine learning innovations on Mac.
