# How to use LM Studio: run AI models on your own computer

> LM Studio is a free desktop app for running open AI models on your own computer. It has a built-in model search, a chat window you can attach files to, and a local server that speaks the OpenAI API. Install it, search for a model, download an option that fits your memory, load it in the Chat tab and ask a question.

Mindket editors · Published 2026-11-06 · Updated 2026-11-06 · https://mindket.com/learn/tools/lm-studio

## In short

- LM Studio runs open AI models on your own computer, with a chat window, a model search and a local server.
- The app costs nothing, and chats and attached files can stay on your machine once a model is downloaded.
- Pick a download option that fits your memory, and aim for 4-bit or higher.
- Existing OpenAI API code can call the local server at port 1234.
- Turn on authentication before you share the server on a network.

## What LM Studio is

LM Studio is a desktop app for downloading and running large language models on your own computer. Element Labs makes it, and it launched in May 2023. Chats and attached documents stay on your machine. Only the model search, the downloads and the updates reach out to the internet.

It runs on macOS with an Apple silicon chip, on Windows and on Linux. It runs models in the GGUF format through llama.cpp, and on Apple silicon Macs it can also run MLX models, Apple's own format. The app is free to download and use. Beside the chat window it comes with a local server, a command line called lms, and Python and TypeScript kits for developers.

## What LM Studio is good for

- **Chatting with open models, offline**: download a model such as Llama, Qwen, Gemma or gpt-oss once, then talk to it with no account and no internet.
- **Answering questions about your own files**: attach a .docx, .pdf or .txt file to a chat and the model answers from it, with the file staying on your computer.
- **Serving an API for your own code**: switch on the server, and code written for the OpenAI API can talk to a local model at http://localhost:1234/v1.
- **Comparing models quickly**: the built-in search pulls from Hugging Face and shows each model in several download sizes, so you can start with a small one.
- **Giving a model tools**: add MCP servers, the standard way to connect tools to a model, and the model can call them. Install only servers you trust.
- **Running from a terminal**: the lms command loads and unloads models, starts the server and streams the log, which suits scripts and machines with no screen.

## When to use LM Studio, and when to pick another tool

LM Studio suits people who want a window for local models. [Ollama](https://mindket.com/learn/tools/ollama) runs the same kind of models from the command line, and [Hugging Face](https://mindket.com/learn/tools/hugging-face) is where the models themselves are published. For the strongest answers, use a cloud assistant such as [ChatGPT](https://mindket.com/learn/tools/chatgpt) or [Claude](https://mindket.com/learn/tools/claude).

| Your job | Use LM Studio? | Or try |
| --- | --- | --- |
| Chat privately with a model, with files that stay on your PC | Yes, this is what it is for | Ollama |
| Pick and download models by clicking in a window | Yes | Ollama, if you prefer the command line |
| Build a script or app against a local OpenAI-style API | Yes | Ollama, which also offers an OpenAI-compatible route |
| Get the strongest answers to hard questions | Not usually, a model has to fit in your computer's memory | ChatGPT, Claude or Gemini |
| Write code with an agent in your editor | Not on its own | Cursor, GitHub Copilot or Claude Code |
| Read a model's page or find a new model | Only to download it | Hugging Face |

## How to use LM Studio, step by step

These steps take you from the installer to a first chat with a file attached. Names and shortcuts are the ones LM Studio's docs give.

1. **Check your computer.** LM Studio asks for an Apple silicon Mac on macOS 14 or newer, a Windows PC (x64 with AVX2, or ARM), or Linux with Ubuntu 20.04 or newer. At least 16GB of RAM is recommended.
2. **Install it.** Download the installer for your system from [LM Studio's Downloads page](https://lmstudio.ai/download) and run it. On Linux the download is an AppImage, a single file that runs without an installer.
3. **Open the model search.** Press `Ctrl` + `2`, or `⌘` + `2` on a Mac. Search by a name such as `gemma` or `qwen`, or paste a Hugging Face address.
4. **Choose a download option.** Each model comes in several sizes with names like `Q4_K_M` or `Q8_0`. Pick one at 4-bit or higher that fits your free memory, and download it.
5. **Load the model.** Open the **Chat** tab, open the model loader at the top, select the model and load it. Downloading and loading are separate steps.
6. **Ask a question.** Type in the chat box and press Enter. The model answers on your computer.
7. **Attach a file.** Attach a `.docx`, `.pdf` or `.txt` file and ask about what it says.
8. **Start the server, if your code needs it.** Open the **Developer** tab and switch on **Start server**. Your code can now call `http://localhost:1234/v1`.

1. Press `Ctrl` + `2`, or `⌘` + `2` on a Mac, to open the model search. Then type a model name such as `qwen3` in the **Search models** box.
   ![An illustration of LM Studio's model search with a gold ring around the Search models box holding the word qwen3.](https://mindket.com/blog/screens/lm-studio-1-model-search.webp?v=b31ddae0e953)
2. Open a model and pick a download option such as `Q4_K_M`. Press its **Download** button once the size fits your free memory.
   ![An illustration of LM Studio's model page with a gold ring around the Download button on the Q4_K_M row.](https://mindket.com/blog/screens/lm-studio-2-download-option.webp?v=9e3988a0d053)
3. Open the **Chat** tab and click the model loader at the top, where it says **Select a model to load**. Pick your model and load it.
   ![An illustration of LM Studio's Chat tab with a gold ring around the model loader bar at the top.](https://mindket.com/blog/screens/lm-studio-3-model-loader.webp?v=34231a81cb62)
4. Open the **Developer** tab and switch on **Start server**. Your code can then call `http://localhost:1234/v1`.
   ![An illustration of LM Studio's Developer tab with a gold ring around the Start server switch.](https://mindket.com/blog/screens/lm-studio-4-start-server.webp?v=a82decead736)

*AI-made illustrations of LM Studio's screens, drawn on 6 November 2026 to match these steps. They are not screenshots, so the real app may look different: follow what you see.*

> LM Studio changes menus between versions, so a label may differ from these steps. The tutorials at the top of this page show the steps on screen.

## An example to try

Attach a short .txt or .pdf file to a chat, such as a one-page note of your own, and send this prompt.

````text
Summarize the attached file in five bullet points. Then list two questions it leaves unanswered.
````

- A good answer is five bullets in the file's own terms, not general knowledge about its topic.
- A short file goes into the chat whole, so the model sees all of it. A long file makes LM Studio pick out passages that look relevant, so put the terms you expect in the answer into your question.
- If the reply is slow or the app runs short of memory, load a smaller download option.

## Tips for better results

- Compare file sizes with your free memory before you press Download. Every option is the same model, and the bigger files need more memory to load.
- The context, the text a model can hold at once, takes memory on top of the model file. A long chat or a big document can run a tight machine short.
- When your code sends a `model` field, copy the identifier from the app, or set your own with `--identifier` on `lms load`.
- Run `lms log stream` in a terminal while your code calls the server, to read what went in and out.
- Start with a small model on a laptop. Qwen's team says its Qwen3.5-9B needs about 7GB to run locally, and the download is only part of the memory it uses.

## Keep the server and MCP tools safe

Out of the box the server accepts requests with no key. LM Studio's network page warns that any bind other than 127.0.0.1 exposes the server beyond your computer, and the Serve on Local Network setting does exactly that. Turn on authentication first. Tokens come from the **Manage Tokens** button, and the app shows a new one only once, so copy it straight away.

MCP servers need the same care. LM Studio's MCP page warns that some servers can run arbitrary code, read your local files and use your network.

## YouTube tutorials

- [LM Studio Full Tutorial - The BEST Way to Run Local Models](https://www.youtube.com/watch?v=XpPtfB2pdio) (Tech With Tim)
- [LM Studio Is Getting Insane — Start Using It Now](https://www.youtube.com/watch?v=OOCioZC4tk0) (Bart Slodyczka)
- [LM Studio Is Getting Insane — Master Local AI Now](https://www.youtube.com/watch?v=FHzWptAH9V0) (Eric Tech)
- [LM Studio Tutorial & Review – The Best Way to Run AI Locally](https://www.youtube.com/watch?v=V5Qap-SNyLU) (Zero to MVP)

## Learn more about LM Studio

- [LM Studio tutorial: run local models in a desktop app](https://mindket.com/learn/local-ai/lm-studio-tutorial)
- [Local LLM hardware requirements: how much VRAM a model needs](https://mindket.com/learn/local-ai/local-llm-hardware)
- [Ollama tutorial: run AI models on your own computer](https://mindket.com/learn/local-ai/ollama-tutorial)

## Frequently asked questions

### Is LM Studio free?

Yes. The app costs nothing to download and use, and the open models it runs are free to download. Your computer pays in memory and time.

### Does LM Studio work offline?

Yes, once a model is downloaded. The model runs on your own hardware, so a chat needs no connection. The app goes online to search for models, to download them and to check for updates.

### How much RAM do I need for LM Studio?

LM Studio recommends 16GB of RAM or more on every system. The model must fit too, and a small one such as Qwen3.5-9B needs about 7GB according to Qwen's team. Bigger downloads need more memory.

### What is the difference between LM Studio and Ollama?

The window. Both run GGUF models through llama.cpp and serve an OpenAI-compatible API on your machine. LM Studio puts a model browser and a chat window in front, and Ollama is built around its command line and a library of named tags.

### Can LM Studio use MCP servers?

Yes, since version 0.3.17, when it became an MCP host. Open the Program tab in the right-hand sidebar, choose Install, then Edit mcp.json, and add servers in the mcpServers format.

### Where does LM Studio keep downloaded models?

In its models directory, which you can see and change under My Models. Models you fetch with `lms get` go to the same folder.

## Sources

- [LM Studio Docs: Welcome to LM Studio](https://lmstudio.ai/docs/app) (read 2026-11-05)
- [LM Studio Docs: System requirements](https://lmstudio.ai/docs/app/system-requirements) (read 2026-11-05)
- [LM Studio Docs: Get started](https://lmstudio.ai/docs/app/basics) (read 2026-11-05)
- [LM Studio Docs: Download an LLM](https://lmstudio.ai/docs/app/basics/download-model) (read 2026-11-05)
- [LM Studio Docs: Chat with documents](https://lmstudio.ai/docs/app/basics/rag) (read 2026-11-05)
- [LM Studio Docs: Offline operation](https://lmstudio.ai/docs/app/offline) (read 2026-11-05)
- [LM Studio Docs: Use MCP servers](https://lmstudio.ai/docs/app/mcp) (read 2026-11-05)
- [LM Studio Docs: Local server](https://lmstudio.ai/docs/developer/core/server) (read 2026-11-05)
- [LM Studio Docs: Authentication](https://lmstudio.ai/docs/developer/core/authentication) (read 2026-11-05)
- [LM Studio Docs: Serve on Local Network](https://lmstudio.ai/docs/developer/core/server/serve-on-network) (read 2026-11-05)
- [LM Studio Docs: lms CLI](https://lmstudio.ai/docs/cli) (read 2026-11-05)
- [LM Studio blog: MCP in LM Studio](https://lmstudio.ai/blog/lmstudio-v0.3.17) (read 2026-11-05)
- [Hugging Face Docs: GGUF](https://huggingface.co/docs/hub/gguf) (read 2026-11-05)
- [Qwen on X: Qwen3.5-9B is now available on LM Studio](https://x.com/Alibaba_Qwen/status/2028664203872251943) (read 2026-11-05)
- [Wikipedia: LM Studio](https://en.wikipedia.org/wiki/LM_Studio) (read 2026-11-06)
- [Tech With Tim on YouTube: LM Studio Full Tutorial](https://www.youtube.com/watch?v=XpPtfB2pdio) (read 2026-11-06)
- [Bart Slodyczka on YouTube: LM Studio Is Getting Insane, Start Using It Now](https://www.youtube.com/watch?v=OOCioZC4tk0) (read 2026-11-06)
- [Eric Tech on YouTube: LM Studio Is Getting Insane, Master Local AI Now](https://www.youtube.com/watch?v=FHzWptAH9V0) (read 2026-11-06)
- [Zero to MVP on YouTube: LM Studio Tutorial and Review](https://www.youtube.com/watch?v=V5Qap-SNyLU) (read 2026-11-06)

[Start the lesson](https://mindket.com/learn/local-ai/lm-studio-tutorial)