Skip to content

THE FIELD JOURNAL

How to Run Local AI on a Mac: Ollama, LM Studio and MLX

By Local AI Scout · Beginner guide

Paper-cut illustration of an open silver laptop on a desk beside a small chip and leaf, a coffee cup and a plant, with a paper cloud kept at a distance

By Local AI Scout · Local · Mac · Guide · Docs-only · Checked November 17, 2026

Short answer: On an Apple Silicon Mac (M1 or newer), install LM Studio or Ollama, download one small model, and chat offline. Both need macOS 14 Sonoma or newer, according to their docs. LM Studio runs both GGUF and MLX models; MLX is Apple's own machine-learning framework for Apple silicon. If you like the command line, Apple's mlx-lm package adds mlx_lm.chat in a few commands. Intel Macs are limited: Ollama runs on the CPU only, and LM Studio doesn't support them. LM Studio recommends 16GB of memory or more. We haven't benchmarked these apps on a Mac for this page.

This guide follows one reader: a freelance bookkeeper with a 16GB MacBook Air who wants a private assistant for client emails and has never opened Terminal.

What your Mac needs

LM StudioOllamamlx-lm
ChipApple Silicon (M1/M2/M3/M4 listed)Apple M series (CPU and GPU); Intel x86 CPU onlyApple silicon
macOS14.0 or newerSonoma (14) or newerNot stated in the README we opened ‡
Memory16GB+ recommended; 8GB possible with smaller models and modest contextModels can need "tens to hundreds of GB" of diskDepends on the model
InterfaceApp with chat windowApp plus Terminal commandTerminal and Python

Apple Silicon Macs share one pool of memory between the processor and graphics, which is why the MLX README describes a "unified memory model." To estimate whether a model fits, use our memory formula.

Path 1: LM Studio (easiest)

  1. Download LM Studio for Mac from lmstudio.ai and open it.
  2. In Discover, pick a small model. On a Mac you'll often see an MLX version beside the GGUF one; either works. LM Studio's command-line tool can filter for MLX downloads with lms get <model> --mlx.
  3. Load it and chat. LM Studio's docs say chatting, chatting with documents and its local server work offline once the model is downloaded.

Our full LM Studio guide covers settings and documents. Not sure which file to pick? What is GGUF? explains the names.

Path 2: Ollama (best for other apps)

  1. Download ollama.dmg, open it and drag Ollama into Applications. On first launch it offers to link the ollama command in /usr/local/bin.
  2. Open Terminal and run ollama run gemma4 (the example in Ollama's CLI docs), or pick a smaller model from our best Ollama models list.
  3. Models live in ~/.ollama. Ollama's macOS page lists every folder to delete if you uninstall; our delete Ollama models guide walks through freeing space.

Our news page reported that Ollama 0.40.0 (September 25, 2026) made MLX the default for supported models on Apple Silicon, per Ollama's release notes. Which models count as "supported" changes by release ‡.

Path 3: MLX from the command line

For people comfortable with Python:

pip install mlx-lm
mlx_lm.chat

The mlx-lm README says mlx_lm.chat opens a chat that keeps context while it runs, mlx_lm.generate --prompt "..." answers one prompt, and mlx_lm.server runs a local server. It pulls models from Hugging Face; the README's examples use the mlx-community collection, such as mlx-community/Llama-3.2-3B-Instruct-4bit. It can also install with conda: conda install -c conda-forge mlx-lm.

Prefer the engine underneath many apps? See How to install llama.cpp; its docs list brew install llama.cpp for Mac.

What about an Intel Mac?

Ollama runs on Intel Macs using the CPU only, so expect slow replies with anything but small models. LM Studio's requirements page says Intel Macs aren't currently supported. For a side-by-side with Windows and Linux, see Local AI on Windows, Mac or Linux.

Choose LM Studio if…

  • You want a chat window and no Terminal.
  • You want to try MLX and GGUF versions of the same model.

Choose Ollama if…

  • You want other apps, like a browser front end or a coding tool, to use your models. Compare both in LM Studio vs Ollama.

Choose mlx-lm if…

  • You already use Python and want Apple's own framework with nothing in between.

Frequently asked questions

Can I run a local LLM on an Intel Mac?

Partly. Ollama's macOS docs say Intel (x86) Macs run on the CPU only, so stick to small models. LM Studio's requirements page says Intel Macs aren't currently supported.

How much memory does a Mac need for local AI?

LM Studio recommends 16GB or more and says 8GB Macs can work with smaller models and modest context sizes. The exact need depends on the model; use a memory formula before downloading.

What is MLX?

MLX is an array framework for machine learning on Apple silicon, published by Apple's ml-explore team. LM Studio can run MLX models on Apple Silicon Macs, and the mlx-lm package runs them from the command line.

Does local AI on a Mac work offline?

Yes, once the model is downloaded. LM Studio's docs say chatting and chatting with documents don't need internet. Searching for and downloading models, and update checks, do.

Sources

  • Ollama Docs, "macOS" (https://docs.ollama.com/macos) and "CLI Reference" (https://docs.ollama.com/cli), opened November 17, 2026.
  • LM Studio Docs, "System Requirements" (https://lmstudio.ai/docs/app/system-requirements) and "Offline Operation" (https://lmstudio.ai/docs/app/offline), plus the docs index at https://lmstudio.ai/llms.txt (MLX on Apple Silicon; lms get --mlx), opened November 17, 2026.
  • ml-explore/mlx README (https://github.com/ml-explore/mlx) and ml-explore/mlx-lm README (https://github.com/ml-explore/mlx-lm), opened November 17, 2026.
  • Ollama 0.40.0 MLX default: as summarized on our /news/ page from Ollama's release notes; not reopened today ‡.

How we checked: a documentation-based guide, not a hands-on test. Sources opened November 17, 2026 unless stated. ‡ marks something we could not confirm on an official page. See our editorial policy.

Keep exploring: Apps · Models · All guides

THE SCOUT LETTER

Good finds.
Less noise.

Useful local AI apps, practical guides and the updates worth your time. A thoughtful note in your inbox.

Free to join. Unsubscribe with one click.