apple/coreai-models

Model export recipes, Python primitives, and Swift runtime utilities for on-device AI

What it solves

This project provides a comprehensive toolkit for deploying open-source AI models on Apple silicon (macOS and iOS). It simplifies the process of converting popular models from sources like Hugging Face into a format compatible with Apple's Core AI framework, while providing the necessary runtime utilities to integrate them into applications.

How it works

The toolkit is divided into four main components:

  • Model Export: Provides recipes to convert PyTorch models into standalone .aimodel files.
  • Python Primitives: Offers building blocks for developers to author custom Core AI models in PyTorch.
  • Runtime Utilities: A Swift package that helps developers integrate and run these exported models within iOS and macOS apps.
  • Agent Skills: Plugins for coding agents (such as Claude Code, Codex CLI, and Gemini CLI) that provide expert guidance on deploying, authoring, and compressing models for Apple platforms.

Who it’s for

Developers building on-device AI applications for Apple platforms who need to convert existing open-source models or create custom models optimized for Apple silicon.

Highlights

  • Curated Model Catalog: Includes a registry of supported models with specific export recipes.
  • On-Device Execution: Specifically designed for macOS and iOS 27.0+.
  • Model Compression: Tools for exploring weight compression, quantization, and palettization via coreai-opt.
  • Coding Agent Integration: Dedicated plugins to help AI coding assistants automate the deployment workflow.

Related

  • Project
    apple/coremltools

    A Python package for converting machine learning models from frameworks like PyTorch and TensorFlow into the Core ML format for on-device deployment on Apple devices.

  • Dispatch
    Apple Core AI Framework: On-Device AI Integration and Optimization

    Apple's Core AI framework provides a new way to convert PyTorch models for high-performance execution across CPU, GPU, and the Apple Neural Engine (ANE), shifting the focus toward local, on-device AI.

  • Project
    rudrankriyam/Foundation-Models-Framework-Lab

    A native iOS and macOS workbench for testing and shipping with Apple's Foundation Models framework, featuring a prompt playground, tool integration, and adapter comparison.

  • Project
    john-rocky/coreai-model-zoo

    CoreAI‑Model‑Zoo is a curated, reproducible catalog of Apple Core AI `.aimodel` bundles (LLMs, VLMs, speech‑to‑text, OCR, safety classifiers, etc.). Each entry ships a conversion recipe, parity checks, and one‑line Swift loading snippets via the CoreAIKit library, plus demo Xcode apps. The repo enables on‑device inference on iOS 27/macOS 27 with full verification and community‑contributed ports.

  • Project
    ml-explore/mlx-lm

    A Python package for generating text and fine-tuning large language models on Apple silicon using the MLX framework.