LM Studio
Description
A desktop app for running large language models on your own computer. You search and download open models (Llama, Qwen, gpt-oss, DeepSeek and others, via Hugging Face) inside the app and chat with them in a graphical interface; you can attach documents and work with them offline. It can start a local server compatible with the OpenAI API, has a REST API, Python and TypeScript SDKs and a command-line tool (lms), works as a client for MCP servers, and has a headless version without a window for servers. Models run on llama.cpp (GGUF) and, on Apple Silicon Macs, MLX. It is available for macOS (Apple Silicon only), Windows (x64 and ARM) and Linux. A graphics card is not required, but the docs recommend at least 4 GB of video memory on Windows, 16 GB of RAM and, for x64, a processor with AVX2. It has no model of its own. According to the official blog, it is free to use at home and at work (checked 25 October 2026).
Pros and cons
Pros
- Free for home and work use
- Convenient graphical interface: find, download and chat in one app
- Local server compatible with the OpenAI API, SDKs and CLI
- Works with documents offline and supports MCP
Cons
- The app itself is not open source; use is under its terms
- On Mac only Apple Silicon is supported (no Intel Macs)
Key facts
- Free access
- Yes Free to use at home and at work
- Developer
- LM Studio
- Checked on
- Oct 25, 2026