LM Studio is a free desktop application that enables users to download, run, and interact with large language models (LLMs) entirely on their local machine — no cloud, no subscription, no data leaving the device. It provides a polished GUI for non-technical users while exposing an OpenAI-compatible REST API and developer SDKs (JavaScript and Python) for programmers who want to integrate local models into their own applications. Supporting a wide range of open-weight models — including Qwen 3, Gemma 3, DeepSeek-R1, and gpt-oss — LM Studio handles model download, quantization selection, GPU/CPU offloading, and inference configuration through a single interface. A standout 2026 addition is LM Link, which allows end-to-end encrypted connections to remote LM Studio instances via a Tailscale partnership, effectively turning any machine running LM Studio into a private AI server. The headless variant, llmster, strips the GUI and exposes the same capabilities for server, cloud, and CI/CD environments. LM Studio also ships with MCP (Model Context Protocol) client support, enabling tool-calling workflows with external services. Version 0.4.12 (April 2026) added Qwen 3.6 support, Anthropic-compatible reasoning output parameters, and improved tool-calling reliability for GLM and gpt-oss model families. The platform is free for personal and commercial use under its published terms of service.