Short answer
- Mac with M1 or later: Apple Intelligence is built in and free.
- Want a full chatbot? Install the ChatGPT, Claude or Copilot Mac app (all have free plans).
- Want it private and offline? Run open models with LM Studio or Ollama. Check yours in the checker.
1. Apple Intelligence
Which Macs: any Mac with Apple silicon (M1 or later), on a recent macOS (macOS 27 Golden Gate is current), with Siri and device language set to a supported language. Some newer features need an M3 or later with at least 12 GB of unified memory. Intel Macs don't get it.
Features include Writing Tools (rewrite, proofread, summarize anywhere you type), Image Playground, a smarter Siri (Siri AI is in beta in English), and AI models in Shortcuts. Some server-side features have daily usage limits.
How it uses ChatGPT: it's an optional extension. Turn it on in System Settings → Siri (or Apple Intelligence & Siri) → ChatGPT. Siri and Writing Tools can then hand some requests to ChatGPT. You don't need a ChatGPT account; signing in to a paid account gives more advanced use. You can block it with Screen Time.
2. AI apps for Mac
| App | Free plan | Good for |
|---|---|---|
| ChatGPT for Mac | Yes | All-round help; ChatGPT Work and dots start on desktop |
| Claude for Mac | Yes | Long, careful writing; Claude can also use your computer in Cowork |
| Microsoft Copilot for Mac | Yes (limited) | If you use Word, Excel and Outlook on Mac |
Download from each company's official page (linked in Sources). Requirements are listed there.
3. Local AI on Apple Silicon
Unified memory means the processor and graphics share one pool of RAM. That's great for AI: a 16 GB Mac can use most of its 16 GB for a model, while a PC is usually limited by its graphics card's VRAM. Leave a few GB for macOS and your apps.
| Your Mac's memory | What to expect |
|---|---|
| 8 GB | Small models (around 1–4B), e.g. Qwen 3.5 4B or Gemma 4 E2B (4.6 GB). Fine for simple tasks. |
| 16 GB | The sweet spot: comfortable 7–12B models. LM Studio recommends 16 GB+. |
| 32 GB+ | Bigger, smarter models (around 26–32B). |
Apps: LM Studio (point-and-click; Apple silicon only, macOS 14+) or Ollama (lightweight; plugs into other apps). Both are free. Step by step: 10-minute Mac setup · Ollama vs LM Studio.
Research-based. From Apple's and the apps' official pages, last checked Oct 15, 2026. We haven't tested Apple Intelligence or local AI on a Mac ourselves; our own local AI test was on Linux. The memory table is a rule of thumb based on model download sizes. On Windows instead? See AI on Windows PCs. On iPhone? See AI on iPhone.
Sources
- Apple Support: How to get the next generation of Apple Intelligence (Sep 14, 2026): last checked Oct 15, 2026
- Apple Support: Use ChatGPT with Apple Intelligence on Mac: last checked Oct 15, 2026
- Apple Support: macOS 27 Golden Gate compatible computers: last checked Oct 15, 2026
- ChatGPT desktop apps: last checked Oct 15, 2026
- Claude desktop apps: last checked Oct 15, 2026
- Microsoft: Copilot plans (works on PCs and Macs): last checked Oct 15, 2026
- LM Studio: system requirements: last checked Oct 15, 2026
- Ollama: download (Mac, Windows, Linux): last checked Oct 15, 2026
Prices and plans change often. Always confirm on the official site before you buy.