LM Studio
Desktop app to discover, download and run LLMs locally on your own computer (Windows, macOS, Linux) — offline, with an OpenAI-compatible local server.
Description
What is LM Studio
LM Studio is a desktop application for discovering, downloading and running large language models (LLMs) entirely locally, on your own computer. It runs on Windows (x64/ARM64), macOS (Apple Silicon) and Linux (x64) and lets you use open-weight models such as Llama, Qwen, Gemma, Phi, Mistral, DeepSeek and gpt-oss without sending any data to the cloud.
Inference engines and model formats
The app is built on two engines running under the hood: llama.cpp (for GGUF-format models, on all platforms) and Apple MLX (for MLX-format models on Apple Silicon Macs). It supports quantised models, which makes it possible to run large models on consumer hardware.
Key features
LM Studio provides a graphical chat interface, built-in model search and download (powered by Hugging Face), offline chat with documents (RAG), a local server that exposes models through an OpenAI-compatible API (locally and over the network), tool use (function calling), structured output, and Model Context Protocol (MCP) server support. Official SDKs for Python and TypeScript and the "lms" command-line tool are also available.
Licensing
LM Studio is free — on 8 July 2025 its makers announced that using the app at work no longer requires a commercial license. It is developed by Element Labs, Inc.