Skip to main content

LM Studio Setup

LM Studio provides a desktop app for running large language models locally, with a built-in server API.

Installation​

Download the desktop app from lmstudio.ai for your platform (Windows, macOS, Linux).

Starting the Local Server​

  1. Open LM Studio
  2. Go to the Local Server tab (sidebar icon)
  3. Select a loaded model
  4. Click Start Server

By default the server runs on http://127.0.0.1:1234 with an OpenAI-compatible API.

  • Enable "Allow CORS" and "Network Access" so other tools on your machine can reach the server
  • Adjust the context window — start small and increase until you find the sweet spot between speed and quality
  • GPU offload layers: set to Max to push as much work to your GPU as possible

Finding & Loading Models​

I'm currently running Qwen3.6-27b on a RTX 4090.

If you are using a M-series Mac then you will want to use MLX variant. LM Studio will have a badge on the model if GPU offloading is fully or partially supported.