---
name: Core AI vs Core ML vs MLX
description: >-
  Teach a beginner how to pick Core ML, Core AI, or MLX for Apple on-device
  work from OS target, model type, and whether they are training or shipping.
---
# Core AI vs Core ML vs MLX

Use this course to pick one Apple stack. The reader is new to on-device AI on Apple platforms. Keep the words simple. State the job of each tool. Do not invent APIs or a deprecation that Apple has not announced.

## End-to-end agent recipe

1. Start with [lesson 01](01-three-names-one-job.llms.md). Separate the three names. Say why an Edge FDE must pick one stack that matches the app.
2. Use [lesson 02](02-core-ml-when.llms.md) for Core ML, `.mlmodel`, coremltools, older OS targets, and classic ML.
3. Use [lesson 03](03-core-ai-when.llms.md) for Core AI, `.aimodel`, coreai-torch, iOS or macOS 27 or later, and Xcode 27.
4. Use [lesson 04](04-mlx-when.llms.md) for MLX as Mac training and research. Then export before shipping.
5. Use [lesson 05](05-choose-with-confidence.llms.md) for the decision table, the checklist, and the sibling-course links.
6. Keep these links exact: [Model architectures in plain English](../model-architectures-plain/), [Core AI models, typed](../coreai-models-typed/), and [PyTorch to Core AI in Xcode](../pytorch-core-ai-xcode/).
7. Cite only [developer.apple.com/core-ai](https://developer.apple.com/core-ai/), [apple/coreai-models](https://github.com/apple/coreai-models), and [MLX](https://github.com/ml-explore/mlx) when a source is needed.
8. Mention Foundation Models only as Apple's system language-model API. It is not the path for a model file the reader trained.
9. Use the human HTML pages for the reader. Use the `.llms.md` briefs for agent steps and acceptance checks.

## Constraints

- Plain spoken English for a non-native reader.
- Literal sentences. No metaphor, no flourish, no quizzes.
- Do not use em dashes.
- Use no Mermaid. Build progressive `.flow` diagrams in HTML.
- Do not invent deprecation. Core ML is still maintained.
- Do not invent OS versions. Core AI needs 27 or later and Xcode 27.
- Keep the sibling links exactly as `../pytorch-core-ai-xcode/`, `../coreai-models-typed/`, and `../model-architectures-plain/`.

## Key concepts

- Core ML, Core AI, and MLX are three Apple tools. They are not the same stack.
- Core ML ships classic models in apps. Core AI ships custom neural or generative models on OS 27 or later.
- MLX trains and tests on a Mac. You export before you ship.
- Foundation Models is Apple's system language-model API. It is not a path for your own trained file.

## Takeaways

- You can pick one stack from OS target, model job, and whether you are training or shipping.
- You can say one job for each name without swapping them.
- You know the sibling courses for shapes, catalog presets, and the PyTorch to Core AI build.

## Acceptance checks

- Every human lesson HTML has **Key concepts** and **Takeaways** sections before lesson-nav. The course home has the same two headings for the whole course.
- Five human lessons exist with the requested basenames.
- Five matching `.llms.md` briefs exist and link to their human HTML pages.
- Every human lesson has its Agent brief link and the three sibling-course links.
- The course index lists all five lessons, the course SKILL.md, and all five briefs.
- Diagrams are progressive HTML `.flow` elements. There is no Mermaid, no em dash, and no quiz.

## Environment setup

No runtime environment is required. The short code samples are sketches of public APIs, not a project to run.