How to use LM Studio on a Mac: install, pick a model, and chat

LM Studio runs AI models on your Mac for free, with no account and no internet once a model is downloaded. You need a Mac with Apple silicon and macOS 14 or later; LM Studio recommends 16 GB of memory, and 8 GB works with small models. Install it, open Discover, download a 4-bit model that fits your memory, and chat. This guide covers each step, plus MLX vs GGUF, documents, the local server, privacy and uninstalling.

Facts come from LM Studio's own documentation, changelog and privacy policy, checked on 19 October 2026, when the current version was 0.4.26. Comparing tools? Read LM Studio vs Ollama first.

Before you start: what you need

• A Mac with Apple silicon (M1 or newer). "Intel-based Macs are currently not supported."
• macOS 14 Sonoma or later. LM Studio says "macOS 14.0 or newer is required."
• Memory: 16 GB or more recommended. On 8 GB, LM Studio says to "stick to smaller models and modest context sizes."
• Disk space for models: 2 to 20 GB each.
• The right app. LM Studio now has two products: the classic LM Studio for running models, and Bionic, a separate agent app with paid cloud plans. This guide is about the classic app. Local use is free, at home and at work.

How to install LM Studio on a Mac

1. Go to lmstudio.ai and download the Mac version.
2. Open the download and drag LM Studio into your Applications folder.
3. Open it from Applications. If macOS asks whether you're sure, choose Open.
4. You don't need a paid plan to run models on your Mac; the free version covers it.

LM Studio updates itself from inside the app when it's online.

Download your first model

1. Open the Discover tab (Command-2) and search for a model, such as Qwen3, Gemma or gpt-oss. You can also paste a Hugging Face link.
2. Pick a version. Labels like Q4 or 4-bit mean the model is compressed; LM Studio's advice is "choose a 4-bit option or higher if your machine is capable enough."
3. Download it, then go to the Chat tab, choose the model from the loader at the top and start typing.

MLX or GGUF? LM Studio runs two kinds of model files. GGUF files run through the llama.cpp engine on any computer. MLX files run through Apple's MLX framework and only work on Apple silicon. On a Mac you can use either; LM Studio doesn't publish a speed comparison, so if both versions of a model exist, try both and keep the faster one.

Which model fits your Mac

A 4-bit model needs roughly its file size in free memory, plus room for the conversation. This table is our estimate from download sizes, not LM Studio's official guidance:

If a model is slow or won't load, pick a smaller one or a smaller quantization before changing settings. For how small models compare in real answers, see our five-model test on a Mac.

Chat with documents

You can attach Word, PDF and text files to a chat. LM Studio decides how to use them: if a document fits in the model's context, "LM Studio will add the file contents to the conversation in full." If it's longer, it switches to retrieval, pulling out the parts that seem relevant to your question. LM Studio notes that retrieval "sometimes requires some tuning and experimentation," so specific questions work better than broad ones.

It works one chat at a time with the files you attach. It doesn't keep a whole folder searchable or update when files change.

Settings worth knowing

• Modes: User mode shows just the chat and sets things up for you. Developer mode, in Settings, shows every option, the server and the command line.
• Context length: how much text the model can hold at once. Bigger uses more memory. Raise it only for long documents.
• Per-model defaults: in My Models, the gear icon saves settings for a model, so you don't set them every time.
• Your chats are saved on your Mac as files you can find with Reveal in Finder. The model doesn't learn from them.

Use LM Studio with other apps

• Local server: in the Developer tab, turn on Start server. Apps that speak OpenAI's or Anthropic's API can then use your model at http://localhost:1234. A setting lets other devices on your network connect too.
• Command line: LM Studio ships a tool called lms. lms ls lists your models, lms load loads one, and lms server start starts the server.
• Without the app window: llmster runs LM Studio as a background service with no interface, for servers or always-on setups.
• MCP tools: LM Studio can connect models to MCP servers. Its own warning: "Never install MCPs from untrusted sources," because some can run code and read your files.

Is LM Studio private? What needs internet

LM Studio says it "can operate entirely offline, just make sure to get some model files first," and that "all document processing is done locally." Its privacy policy says none of your messages, chat histories or documents "are ever transmitted from your system," and that the app has no telemetry or user tracking.

Some things still go online: searching and downloading models, and update checks, which send the app version, operating system and your IP address. LM Studio also collects anonymized search terms and download counts. Optional cloud features and LM Link, for using a Mac's models from another device, use the internet. Chatting with a downloaded model doesn't.

How to uninstall LM Studio

LM Studio doesn't document an uninstall. Dragging the app to the Trash leaves the models behind, and they're the big part.

1. Delete models you don't want in My Models first, or note where your models folder is in its settings.
2. Quit LM Studio and stop the server if it's running (lms server stop).
3. Move LM Studio from Applications to the Trash.
4. In Finder, press Shift-Command-G and go to ~/.lmstudio. That folder holds models and chats by default; delete it only if you don't want them back.

Common problems

• The model won't load or is very slow: it's too big for your memory. Choose a smaller model or quantization, or lower the context length.
• Search shows nothing: Discover needs internet.
• Another app can't connect: check the server is on and the port is 1234, or change the port if something else uses it.
• Document answers miss things: long files go through retrieval, which only sees some parts. Ask about one specific point at a time.
• It won't install: check you have Apple silicon and macOS 14 or later.

If your questions are about whole folders

LM Studio is the best free way to chat with models and run a local server. If you mostly ask about your own documents, Citenook is a Mac app built for that: it keeps whole folders of PDFs, Word files and notes searchable, updates as files change, and links every answer to the exact passage. It can run its own downloaded models or, on the Pro plan, use the models you already have in LM Studio. It's not a general chat app, it's Mac only, and it costs from $9.99 a month after a 30-day free trial.

Questions

Sources, checked 19 October 2026: LM Studio system requirements, downloading models, chat with documents, offline operation, MCP, lms CLI, headless mode, changelog, pricing, free for work and privacy policy; MLX in LM Studio; model sizes from Ollama's library, Hugging Face model cards and Google's Gemma docs.

Related: LM Studio vs Ollama · Offline AI apps and models · Can you use ChatGPT offline?