LMStudio
LM Studio is a user-friendly desktop application designed for experimenting with local and open-source Large Language Models (LLMs). It all...
Last verified:
What is LMStudio?
LM Studio is a desktop application that lets you discover, download, and run large language models (LLMs) locally on your computer, entirely privately without sending data to the cloud. It supports popular open-source models like Llama, Qwen, Gemma, DeepSeek, Phi, and Mistral in GGUF and MLX formats. The app provides an intuitive graphical interface for browsing models via Hugging Face, managing downloads, and chatting with any downloaded model.
Key features include a flexible chat interface for interacting with models, document chat (RAG) for querying your files offline, MCP server integration for tool calling and agent workflows, and a local REST API server with OpenAI-compatible and Anthropic-compatible endpoints. LM Studio also offers SDKs for TypeScript and Python, a CLI tool (lms), structured JSON output via schema enforcement, embeddings generation, and LM Link for routing workloads across devices on a private network. It supports GPU offloading including integrated GPUs for faster inference.
LM Studio is designed for privacy-conscious users, developers building local AI applications, hobbyists experimenting with open-source LLMs, researchers, and businesses that need data to stay on their own hardware. It runs on macOS (including Apple Silicon), Windows (x64/ARM64), and Linux (x64), with a headless version called llmster for server/CI deployments without a GUI. The tool is ideal for offline AI use, local development, coding assistants, and building custom AI workflows without cloud reliance.
The application emphasizes complete data privacy with no data collection, no telemetry, and no ads. It has over 9.9k followers on GitHub and millions of downloads worldwide, with Fortune 500 companies, universities, and global organizations using it for enterprise deployments. Users can run models entirely offline, attach documents to chat messages for RAG, and serve models on network-accessible endpoints for other tools to call.
LMStudio pricing
Pricing model: Free
LM Studio is completely free for both personal and business use with no subscription fees, trial required, or hidden costs. As of July 2025, the commercial license requirement for work use was removed - teams can use it at home, work, school, or anywhere for free. All features are included at no charge. An Enterprise plan is available for companies requiring advanced features like SSO, model/MCP gating, and private collaboration (contact required). A self-serve Teams plan for private artifact sharing within teams was announced as coming later in July 2025. The LM Studio Hub allows creating public organizations for teams to use for free.
LMStudio pros
- Completely free for personal and business use with no subscription fees
- Runs AI models entirely locally with complete data privacy and no cloud dependency
- No data collection, no telemetry, no ads - excellent privacy standards
- Intuitive graphical UI for browsing, downloading, and managing models
- Built-in search and download functionality via Hugging Face integration
- Supports multiple model formats including GGUF (llama.cpp) and MLX (Apple Silicon)
- OpenAI-compatible and Anthropic-compatible API endpoints for easy integration
- Local REST API v1 with stateful chat, completions, and embeddings endpoints
- TypeScript SDK (lmstudio-js) and Python SDK (lmstudio-python) for developers
- CLI tool (lms) for scripting downloads, daemon management, and server control
- MCP server integration for tool calling and local agent workflows
- Structured JSON output via JSON schema enforcement for typed responses
- Document chat (RAG) for querying your files entirely offline
- GPU offloading support including integrated GPUs for faster inference
- LM Link for routing workloads across devices on encrypted private network
- Headless mode (llmster) for server/CI deployments without GUI
- Cross-platform support for macOS (Apple Silicon), Windows, and Linux
- Model management features including load, unload, list, and configuration
- Chat interface works with any downloaded model for immediate use
- Active community on Discord with direct team support and regular updates
LMStudio cons
- Model inference speed depends heavily on your local hardware capabilities
- Some models below 7B parameters may not support structured JSON output
- Requires downloading large model weights before first use (storage intensive)
- Learning curve for exploring all features and developer integrations
- May lack certain niche features that specialized competitors offer
- No built-in model fine-tuning capabilities - only inference support
- Limited to open-source models available on Hugging Face ecosystem
- GPU offloading may not work optimally on all integrated GPU configurations
- Not suitable for collaborative cloud-based features or multi-user cloud access
Frequently asked questions about LMStudio
Is LM Studio completely free to use?
Yes, LM Studio is completely free to use with all features included. There are no paid tiers, subscription fees, or hidden costs. As of July 2025, it is free for both home and work use - the commercial license requirement was removed. You and your team can just use the app without filling out forms or contacting the company.
What models can I run with LM Studio?
LM Studio supports running local LLMs like gpt-oss, Llama, Qwen3, Gemma3, DeepSeek, Phi, Mistral, and many more. It supports GGUF format (via llama.cpp) on Mac, Windows, and Linux, and MLX format on Apple Silicon Macs. You can search and download models directly within the app via Hugging Face integration.
Does LM Studio send my data to the cloud?
No, LM Studio runs entirely locally on your computer with complete data privacy. Your data never leaves your machine - there is no internet capability with the models themselves, no data collection, no telemetry, and no ads. This makes it ideal for privacy-conscious users and businesses that need data to stay on their own hardware.
Can I use LM Studio offline?
Yes, you can chat with documents and interact with models entirely offline on your computer. After downloading model weights once, all inference happens locally. The document chat feature (RAG) lets you attach documents to chat messages and query them without any internet connection.
What operating systems does LM Studio support?
LM Studio is available for macOS (including Apple Silicon), Windows (x64 and ARM64), and Linux (x64). You can download installers for your specific operating system from the Downloads page. Apple Silicon Macs additionally support running MLX models.
How do I integrate LM Studio with my own applications?
LM Studio provides a v1 REST API with OpenAI-compatible and Anthropic-compatible endpoints that you can use from your own apps and scripts. It also offers TypeScript SDK (lmstudio-js) and Python SDK (lmstudio-python) for building local AI workflows. You start the server from the Developer tab or via the lms CLI.
What is LM Link and what does it do?
LM Link routes local AI workloads across devices and preferred machines on your network. Any model in your LM Link network can be used as if it were local, even if it's on a remote device. All devices communicate using Tailscale's end-to-end encrypted connections and are never exposed to the public internet. It is entirely separate from other Tailscale usage and won't interfere.
Can I deploy LM Studio on servers without a GUI?
Yes, llmster is the headless version of LM Studio without the desktop GUI. It's ideal for servers, CI environments, or any machine where you don't need a graphical interface. You can install it using curl on Mac/Linux or PowerShell on Windows, and it runs standalone as a daemon.
Does LM Studio support structured JSON output?
Yes, LM Studio supports structured JSON output through JSON schema enforcement via the /v1/chat/completions endpoint. When given a JSON schema, the LLM will respond in valid JSON conforming to that schema, following OpenAI's Structured Output API format. However, not all models below 7B parameters are capable of structured output.
What developer tools does LM Studio offer?
LM Studio offers a comprehensive developer stack including: TypeScript SDK (lmstudio-js), Python SDK (lmstudio-python), v1 REST API with stateful chat/completions/embeddings, OpenAI-compatible and Anthropic-compatible endpoints, CLI tool (lms) for chat/downloads/daemon management, MCP client for tool calling, structured output via JSON schema, embeddings and tokenization APIs, and model management (load/download/list) functions.