LM Studio is a free desktop app to run local LLMs on your own computer — a private, offline ChatGPT alternative for macOS, Windows, and Linux. Download open models like gpt-oss, Llama, Qwen, DeepSeek R1, Mistral and Gemma from Hugging Face, chat with your documents, and serve an OpenAI-compatible local API. To get started, head over to the Downloads page and grab the installer for your operating system.
Get started with LM Studio #
System requirements #
LM Studio generally supports Apple Silicon Macs, x64/ARM64 Windows PCs, and x64 Linux PCs.
| Platform | Chip / CPU | OS | Memory |
|---|---|---|---|
| macOS | Apple Silicon (M1/M2/M3/M4) — Intel Macs are not supported | macOS 14.0+ | 16GB+ RAM recommended |
| Windows | x64 (AVX2 required) or ARM (Snapdragon X Elite) | Windows 10/11 | 16GB RAM · 4GB+ VRAM recommended |
| Linux | x64 (AVX2) or ARM64 (aarch64) | Ubuntu 20.04+ · AppImage | 16GB RAM · 4GB+ VRAM recommended |
What can I do with LM Studio? #
- Download and run local LLMs like
gpt-oss,Llama,Qwen,MistralorDeepSeek R1— private AI on your own hardware, no cloud account needed - Use a simple and flexible chat interface — a private AI chatbot with system prompts, presets and streaming replies
- Connect MCP servers and use tools with local models — files, browsers, and your own automations
- Search & download models via Hugging Face 🤗 — GGUF and MLX weights, managed in one library
- Serve local models on OpenAI-like endpoints , locally and on the network — plus an Anthropic-compatible Messages API for Claude Code
- Manage your local models, prompts, and configurations — GPU offload, context length, and per-model presets
Popular use cases #
Why people run LLMs locally with LM Studio instead of renting a cloud chatbot.
Run llama.cpp (GGUF) or MLX models #
LM Studio supports running LLMs on Mac, Windows, and Linux using llama.cpp. On Apple Silicon Macs, LM Studio also supports running LLMs using Apple’s MLX framework.
To install or manage LM Runtimes, press ⌘ Shift R on Mac or Ctrl Shift R on Windows/Linux.
Run an LLM like gpt-oss, Llama, Qwen, Mistral, or DeepSeek R1 on your computer #
To run an LLM on your computer you first need to download the model weights. You can do this right within LM Studio — open the Discover tab, search Hugging Face, and pick a size your machine can hold.
lms get qwen3
lms load openai/gpt-oss-20b --gpu=max
lms chat
Offline operation #
LM Studio can operate entirely offline — just make sure to get some model files first. Chats, inference, and local APIs stay on your machine.
Developer — local REST API #
LM Studio provides a REST API that you can use to interact with local models from your own apps and scripts. Serve OpenAI-compatible endpoints (/v1/chat/completions, responses, embeddings) and an Anthropic-compatible Messages API — locally or on your network.
lms server start
curl http://localhost:1234/v1/models
For servers, cloud instances, and CI, install llmster — LM Studio's headless daemon, no desktop app required:
curl -fsSL https://lmstudio.ai/install.sh | bash # macOS / Linux
irm https://lmstudio.ai/install.ps1 | iex # Windows
Python SDK #
Use lmstudio-python to load models, generate text, embed content, and build agents from scripts, notebooks, and backend services.
pip install lmstudio
TypeScript SDK #
Use lmstudio-js to build apps, tools, and local AI workflows in Node.js or TypeScript — including plugins.
npm install @lmstudio/sdk
CLI — lms #
lms ships with LM Studio, so no extra install is needed. It is MIT licensed and developed in the open on GitHub. Bootstrap it once, then automate your local LLM workflows.
~/.lmstudio/bin/lms bootstrap
cmd /c %USERPROFILE%/.lmstudio/bin/lms.exe bootstrap
| Command | What it does |
|---|---|
lms chat |
Chat with a model interactively in the terminal |
lms get |
Search and download models |
lms ls |
List models on disk (mirrors the My Models tab) |
lms ps |
List models currently loaded in memory |
lms load |
Load a model — options like --gpu=max ,--context-length ,--identifier |
lms unload |
Unload a model ( --all for everything) |
lms server start |
Control the local server |
lms runtime |
Manage and update inference runtimes |
lms daemon |
Manage the headless background daemon |
lms link |
Manage LM Link across devices |
Bionic #
Bionic is LM Studio’s agent for open models — natively local, built for creativity, work, and code. It excels at documents, coding tasks, automations, and real-time voice transcription, all processed on-device. For demanding tasks, run Bionic with frontier open models in a Zero Data Retention cloud.
Integrations #
Connect LM Studio to the tools you already use. With the Anthropic-compatible /v1/messages endpoint, Claude Code can run on your local models with a single base-URL change:
export ANTHROPIC_BASE_URL=http://localhost:1234
export ANTHROPIC_AUTH_TOKEN=lmstudio
claude # now running on your local model
Also works with Codex, VS Code, MCP tools, and any OpenAI-compatible client.
LM Link #
LM Link securely connects models across your devices — route local AI workloads to preferred machines and use remote models from another device as if they were local.
Changelog #
Current release: 0.4.23. Highlights of the 0.4.x line include the Anthropic-compatible Messages API, llmster headless daemon, and LM Link. Beta and experimental builds are published on the beta-releases page.
Frequently asked questions #
lms get and lms chat from the terminal. No cloud account needed.