The ButlerUseful technology. Clearly explained.
Home / AI setup guides
The Butler field guides

LM Studio: your first local AI chat

Install a desktop app, download a model that fits your computer, and try a private local conversation without starting with a terminal.

Documentation checked November 14, 2026. Installation paths have not all been hands-on tested. Follow the linked official requirements if they change.

Start here

Choose LM Studio if you want to try an AI model on your own computer with a graphical interface. The app manages model downloads and chats; the downloaded model supplies the answers. Start with one small model and one ordinary task, not a collection of enormous downloads. This guide was checked against official documentation on November 14, 2026; it is not a hands-on installation test.

1. Check your computer

  • Mac: Apple Silicon and macOS 14 or newer. Intel Macs are not supported by LM Studio. The developer recommends 16GB or more RAM; an 8GB Mac needs a smaller model and modest context.
  • Windows: choose the correct x64 or ARM build. x64 needs AVX2 support. Recommended resources include 16GB RAM and 4GB dedicated graphics memory.
  • Linux: AppImage builds support x64 and ARM64. The requirements page lists Ubuntu 20.04 or newer, while warning that versions newer than 22 are less tested.

These are app recommendations, not a promise that every model will fit. Model weights, conversation length and other running apps all consume memory. Official hardware requirements.

2. Install and load one model

  1. Visit the official download page, find the Download LM Studio section (not Bionic or the headless llmster installer), select your operating system and install the downloaded app. On Linux, use your file manager's permissions to allow the trusted AppImage to run.
  2. Open Discover. Choose a small curated model appropriate to your hardware. Read its license and download size before starting.
  3. Go to Chat, open the model selector, and load the downloaded model. A download on disk is not the same as a model loaded into memory.
  4. Ask: Give me three practical uses for a local AI assistant, in one sentence each. Success means an answer appears in the conversation without an API-key setup.

Official first-chat walkthrough. Different download variants trade size against fidelity; start conservatively rather than assuming the largest file is best. You can change the model-storage directory through My Models. Model-download guidance.

3. Try a document, then check the answer

Attach a short, non-sensitive TXT, PDF or DOCX containing something you can verify, such as a meeting agenda. Ask which three items need preparation. Short documents may fit entirely into the model's context; long documents use retrieval to select relevant passages. Ask a specific question and compare the answer against the original rather than treating a confident response as proof. Document-chat guidance.

Privacy, updates and common problems

Downloaded models and built-in document chat can operate offline. Finding models, downloading runtimes and checking updates need internet access. This local workflow does not require buying cloud API usage. External tools or services you later connect need their own privacy review. macOS and Windows include an app updater; use the official download for Linux updates. Offline behavior and updating.

  • Won't load or becomes very slow? Try a smaller model, shorter context and fewer other applications before changing advanced settings.
  • Can't find a model? Check the internet connection and whether its download completed.
  • Before upgrading or uninstalling: save important conversations and keep a separate copy of original documents. Removing a replaceable model download is different from deleting your only copy of work.

Official references

Choose another guide