Lms
LM Studio CLI
Last verified:
What is Lms?
LM Studio is a desktop application that lets you download and run open-source large language models (LLMs) like Llama, Qwen, Gemma, DeepSeek, and gpt-oss directly on your computer privately without internet connectivity. The tool provides an intuitive graphical interface for browsing models from Hugging Face, managing downloaded models, chatting with models, and interacting with your local documents through RAG (retrieval augmented generation) entirely offline.
Key features include a local REST API compatible with OpenAI's format, Python and JavaScript SDKs for developers, GPU acceleration with CPU fallback, support for llama.cpp GGUF models and Apple MLX models on Mac, quantization options for optimizing model size and speed, MCP server integration, LM Link for routing workloads across devices on an encrypted network, and llmster headless mode for server/CI deployment without GUI. The chat interface supports temperature adjustment, conversation history, and document attachment for private document Q&A.
LM Studio is designed for privacy-conscious users, developers building local AI applications, hobbyists experimenting with open-source LLMs, educators, researchers, and professionals who need offline AI capabilities. It works on macOS (Apple Silicon and Intel), Windows (x64/ARM64), and Linux (x64), making it accessible across different hardware configurations for local AI inference without cloud dependency.
Lms pricing
Pricing model: Freemium
LM Studio is completely free to download and use for both personal and internal business/work use. As of July 2025, the commercial license requirement for company use was removed - you can use it at home and at work without any subscription or payment. There are no hidden costs, usage restrictions, or premium features behind a paywall. Users can freely download the app, search and download models from Hugging Face, load them locally, and run them without any fees.
Lms pros
- Completely free for personal and work use with no subscription
- Runs AI models locally with total data privacy - nothing uploaded
- Works offline without internet connection
- Intuitive graphical interface simplifies LLM workflows
- Supports many open-source models: Llama, Qwen, Gemma, DeepSeek, Mistral
- OpenAI-compatible API for easy integration with existing tools
- Python SDK and JavaScript SDK available for developers
- GPU acceleration with CPU fallback for broad hardware support
- Apple Silicon native support with MLX model runtime
- Quantization options to optimize model size and inference speed
- Chat with documents feature (RAG) for private document Q&A
- Cross-platform: Windows, macOS, and Linux supported
- llmster headless mode for server and CI deployment without GUI
- LM Link enables secure multi-device model sharing on encrypted network
- Built-in Hugging Face model search and download functionality
- Hardware detection automatically identifies GPU and system memory
- No data collection and unchanged privacy policy
- Lightweight native app without Electron or external runtime dependencies
Lms cons
- Requires powerful hardware for large models (70B+ parameters)
- No official mobile app version available
- Battery consumption high on laptops during heavy workloads
- Model quality varies depending on which open-source model chosen
- Beginners may struggle with quantization concepts initially
- Not built for high-volume commercial or enterprise use
- Some bugs reported with model loading and occasional crashes
- Advanced fine-tuning options overwhelm non-technical users
- GUI layer adds minor performance overhead compared to CLI-only tools
- Limited collaborative features compared to cloud-based AI platforms
Frequently asked questions about Lms
What is LM Studio?
LM Studio is a desktop application that allows you to download, deploy, and interact with open-source large language models (LLMs) like Llama, Qwen, Gemma, DeepSeek, and gpt-oss directly on your computer. It runs models locally without needing internet or external servers, providing complete privacy and offline functionality. The app includes an intuitive GUI for model management, chat interface, OpenAI-compatible API, and developer SDKs.
Is LM Studio free?
Yes, LM Studio is entirely free to download and use. Starting July 2025, it is free for both personal use and work/commercial use at companies or organizations. There are no subscription fees, hidden costs, usage restrictions, or premium features behind a paywall. You can freely download the app, search for models on Hugging Face, download them, and run them locally on your hardware.
What models does LM Studio support?
LM Studio supports popular open-source LLMs including Llama (various versions), Qwen, Gemma, DeepSeek, Mistral, Phi, OpenOrca, and many more in GGUF format via llama.cpp. On Apple Silicon Macs, it also supports Apple MLX models. You can search and download models directly within the app from Hugging Face. Models range from compact 7B parameter models to larger 70B parameter models.
Does LM Studio work offline?
Yes, LM Studio works completely offline once models are downloaded. Since everything runs locally on your computer, no data is sent to external servers and you don't need an internet connection to use the models. This makes it ideal for secure environments, travel, or locations with unstable internet. You only need internet initially to download models from Hugging Face.
What are the system requirements for LM Studio?
LM Studio supports Apple Silicon Macs (M1/M2/M3/M4/M5), x64/ARM64 Windows PCs, and x64 Linux PCs. It generally requires 16GB RAM minimum for smooth operation with 7B-8B models, though higher specs deliver better performance with larger models. GPU acceleration is supported including integrated GPUs. For large 70B parameter models, high-end hardware with substantial RAM and a powerful GPU is recommended.
How do I use LM Studio's API?
LM Studio provides a local REST API that is OpenAI-compatible. You can serve local models on OpenAI-like endpoints at localhost:1234 by default. Use the Python SDK (lmstudio) or JavaScript SDK (@lmstudio/sdk) to load models, generate text, create embeddings, and build agents. Any tool that connects to a local OpenAI-compatible API can use LM Studio by pointing to localhost:1234.
What is llmster?
llmster is LM Studio's headless daemon version without the GUI. It's designed for deployment on Linux servers, cloud servers, or CI environments where a desktop interface isn't needed. You can install it using curl scripts on Mac/Linux or PowerShell on Windows. llmster supports the same model loading, API serving, and CLI commands as the desktop app but runs entirely in the background.
What is LM Link?
LM Link is a feature that lets you route local AI workloads across multiple devices on your network. It uses Tailscale's end-to-end encrypted connections to connect LM Studio or llmster instances securely. Any model in your LM Link network can be used as if it's local, allowing you to access models on remote devices by pointing to localhost:1234. Your devices are never exposed to the public internet.
Can I chat with my documents using LM Studio?
Yes, LM Studio supports retrieval augmented generation (RAG) which lets you attach documents like PDFs, Word docs, or text files to your chat messages and interact with them entirely offline. This feature allows private document Q&A without uploading your files to any external service. The document chat happens locally on your computer maintaining full privacy.
Does LM Studio support MCP servers?
Yes, LM Studio works as an MCP (Model Context Protocol) client. You can install MCP servers in LM Studio and use them with your local models for enhanced functionality. The app supports integration with Codex, Claude Code, OpenClaw, MCP tools, and remote workflows. If you're developing an MCP server, there's documentation available for adding an 'Add to LM Studio' button.