LM Studio

~/runtime/lm-studio

LM Studio

Run AI models locally and privately. Full GUI plus headless `llmster` for servers.

License

Proprietary (free for home and work use)

Platform

macOS · Linux · Windows

Model formats

GGUF · MLX

API

OpenAI-compatible · JS SDK · Python SDK

What it is.

$ ./vrambudget --runtime lm-studio

Desktop app for browsing, downloading, and chatting with local LLMs. Supports GGUF and MLX formats. Now ships `llmster`, a no-GUI deployment binary for Linux boxes, cloud servers, and CI. Free for home and work use. Strong onboarding for users new to local inference; popular on Apple Silicon thanks to MLX support.

Install.

$ pkg install lm-studio

curl -fsSL https://lmstudio.ai/install.sh | bash

Or download the GUI app from lmstudio.ai (Mac / Win / Linux)

Supported platforms: macOS, Linux, Windows

Features.

$ cat features.md

GUI app

Polished desktop client for Mac (Apple Silicon), Windows, Linux. Model browser, download manager, chat UI, parameter tuning.

llmster headless

Same engine without the GUI. Curl-install on Mac/Linux; PowerShell installer on Windows. Deploy on cloud servers or in CI.

LM Link

Connect to remote LM Studio instances and use their loaded models as if they were local.

SDKs

Official Python (`pip install lmstudio`) and JS (`npm install @lmstudio/sdk`) clients. OpenAI-compatible API surface.

Best for

▸Users who want a GUI for browsing and trying models
▸Apple Silicon (first-class MLX support alongside GGUF)
▸Quick experimentation: download, load, chat in three clicks
▸Headless deployment via `llmster` on servers and CI

Caveats

▸Not open source (proprietary license, free use only)
▸In-memory KV cache; long agentic sessions can churn
▸No native tool-calling parity with the agent-focused servers (oMLX, vLLM)

Discussion.

$ gh discussion list

// sign in with github to leave a comment. threads live in the repo's discussions tab.

What it is.

Install.

Features.

Links.

Compare to…

Discussion.