LM Studio
: How to use it, features, and the business problems it solves
Add bookmark
What is LM Studio?
A desktop app from US-based Element Labs, Inc. that lets you browse open models such as Qwen, Gemma, gpt-oss and DeepSeek on screen, download them to your own computer (Mac, Windows, Linux), and use them for chat or as an OpenAI-compatible local API server. It is free for both personal and work use, and the interface can be switched to Japanese. Its sister AI agent app, "LM Studio Bionic", also offers paid plans for using open models in the cloud (Bionic+ at $20/month, Pro at $100/month).
Business problems it solves
About "LM Studio"
What is LM Studio?
LM Studio is a desktop app that lets you browse open AI models such as Qwen, Gemma, gpt-oss and DeepSeek on screen, download them to your own computer and chat with them right away. It runs on Mac, Windows and Linux, and you can search, download, load and chat with a model entirely from the app's interface, without typing commands. Models come from Hugging Face, and the inference engines are llama.cpp (GGUF format) and, on Apple silicon Macs, Apple's MLX.
The app can also start a local API server (by default at http://localhost:1234) with OpenAI-compatible and Anthropic-compatible endpoints. That means you can point apps written for the ChatGPT API, or tools such as Codex, Claude Code, Cline and Kilo Code, at the models on your own machine. The official documentation states that what you type when chatting with downloaded models, or when chatting with documents, does not leave your computer.
LM Studio is free for both personal and work use (in July 2025, the requirement to obtain a separate commercial license for work use was removed). The company has also released a sister app, "LM Studio Bionic", an AI agent that gets tasks done, and it offers paid plans (Bionic+ and Pro) for using open models in the cloud.
How to use
- Install it Get the installer for your OS from "Download" on the official website. For Linux, both an AppImage and a .deb package are available.
- Find and download a model In the "Discover" tab, pick from the official recommendations or search by keyword such as "qwen" or "gemma". A single model often comes in several versions such as "Q4_K_M" or "Q8_0"; these differ in how much they are compressed (quantization). The official documentation recommends choosing a 4-bit version or higher if your machine can handle it.
- Load a model and chat Select and load a model in the "Chat" tab to talk to it in a ChatGPT-like interface. You can also attach .docx, .pdf and .txt files and ask questions about their contents.
- Use the local API server
Start the server from the developer view and an OpenAI-compatible API becomes available at
http://localhost:1234/v1. Simply change the base URL of the OpenAI SDK to this address to call your local models from existing code. - Work from the terminal (optional)
With the bundled CLI "
lms", you can runlms get(download a model),lms ls(list models),lms ps(loaded models),lms server start(start the server),lms chat(chat in the terminal) and more.
Key features
01Finding and running models
- Model downloads — Search for and download supported models on Hugging Face from within the app. You can also paste a Hugging Face URL to find a model
- Two inference engines — llama.cpp (GGUF) works on Mac, Windows and Linux, and Apple silicon Macs can also run MLX models
- Offline operation — Once a model is downloaded, chatting, chatting with documents and the local server all work without an internet connection
- Parallel requests — The llama.cpp engine can batch multiple requests and process them concurrently (up to 4 by default)
02Chat
- Chat with documents (RAG) — If an attached document is short, the whole text is passed to the model; if it is long, the relevant parts are retrieved. All processing happens on your computer
- MCP support — Add MCP servers (local or remote) so that models can use external tools
- Presets — Save system prompts and generation settings, and import or share them
03For developers
- Local API server — Provides OpenAI-compatible endpoints (`/v1/chat/completions`, `/v1/responses`, `/v1/embeddings`, etc.), an Anthropic-compatible endpoint (`/v1/messages`) and its own REST API. Authentication can be made mandatory
- SDKs and CLI — There are TypeScript (lmstudio-js) and Python (lmstudio-python) SDKs, and the "lms" CLI is published under the MIT License
- llmster — A headless daemon version that runs without a GUI, so you can use LM Studio's capabilities on servers, in the cloud or in CI environments
- LM Link (preview) — Use models running on another computer at home or at work, over an encrypted connection (built on Tailscale), as if they were local
Pricing
LM Studio itself is free, and no sign-up or commercial license is needed for either personal or work use. There is no charge for running models on your computer, including chat and the local server.
Paid plans exist for the sister app, LM Studio Bionic, and apply when you use open models running in the cloud. Prices are in US dollars and are based on the official pricing page as of October 10, 2026.
| Plan | Price | Main features |
|---|---|---|
| Free | $0 | Bionic agent features, running local models (llama.cpp, MLX), offline voice transcription, LM Link (up to 5 devices), limited web search |
| Bionic+ | $20/month | Everything in Free, plus US-hosted open models (Kimi K3, GLM 5.3, DeepSeek V4 Flash and more), discounted bulk tokens, web search and page extraction |
| Pro | $100/month | Everything in Bionic+, plus 5× usage limits and early access to new features |
How credits work: Using cloud models consumes credits based on the model and the number of tokens processed. Usage beyond the allowance included in your plan is covered by purchased credits. Local models and models accessed through LM Link do not consume credits and can be used without an LM Studio account. For teams, organizations can centralize billing for credits, and subscription plans for teams are listed as "coming soon" (at the time of checking).
LM Studio vs. LM Studio Bionic
| LM Studio | LM Studio Bionic | |
|---|---|---|
| Positioning | App for running local LLMs (chat interface and API) | AI agent for open models |
| Main uses | Trying and comparing models, fine-grained settings, serving an API to apps | Creating and editing documents, coding, research, automation |
| Available models | Local models, models via LM Link | Local models, models via LM Link, open models in the cloud |
| Price | Free | Free $0 / Bionic+ $20/month / Pro $100/month |
The official documentation explains that Bionic is a separate app from LM Studio, and that you can keep using LM Studio alongside Bionic for advanced low-level configuration. When Bionic's coding mode is enabled, it can edit files, use Git and run shell commands within the folder you choose.
System requirements
- macOS — Apple silicon (M1–M4) Macs running macOS 14.0 or later. 16GB+ of RAM is recommended (8GB Macs may still work with smaller models and modest context sizes). Intel-based Macs are not supported.
- Windows — Supports x64 and ARM (Snapdragon X Elite). On x64, a CPU with AVX2 support is required. 16GB+ of RAM and a GPU with at least 4GB of dedicated VRAM are recommended.
- Linux — Supports x64 and ARM64. Distributed as an AppImage (a .deb is also available); Ubuntu 20.04 or later is required.
- Model size — A single model can range from a few GB to tens of GB or more. Larger models need more memory.
Japanese support
- App interface — Japanese can be selected under "Language" in the settings (translated by community localizers). To open the settings, you need to be in Power User mode or higher.
- Website and documentation — In English.
- Chatting in Japanese — Whether Japanese works depends on the model. Choose a multilingual model such as Qwen or Gemma to converse in Japanese.
- Pricing (Bionic) — In US dollars; no yen pricing is shown.
Pros and cons
Pros
- Free for personal and work use, and your interactions with downloaded models stay on your computer
- Everything from finding a model to chatting is done in the interface, so it is easy to try local LLMs even if you are not comfortable with the command line
- The app interface can be displayed in Japanese
- OpenAI-compatible and Anthropic-compatible APIs let you point existing apps and coding tools at your local models
- Quantization versions and load settings can be chosen in detail on screen, which makes comparing models easy
Cons
- Running it comfortably requires a computer with plenty of memory or a powerful GPU
- Not available on Intel Macs
- The app itself is not open source (the lms CLI and SDKs are)
- Agent features and cloud models are in a separate app (Bionic), whose paid cloud plans are priced in US dollars and are generally non-refundable
Reputation and reviews
LM Studio is widely known as a GUI app for running local LLMs, and the company's blog says it has been downloaded millions of times worldwide and deployed at dozens of enterprises. On this site, LM Studio also appears as a local model provider on the Cline and Kilo Code pages. Compared with Ollama, which is operated through commands, LM Studio's strength is that finding models, chatting and adjusting settings all happen in the interface, which suits people trying local LLMs for the first time or those who want to compare quantization versions before choosing. On the other hand, like other local runtimes, large models may not run, or may run slowly, depending on your computer's specs.
Frequently asked questions (FAQ)
Q. Is LM Studio free? A. Yes. LM Studio itself is free for both personal and work use. Paid plans exist for the sister app LM Studio Bionic, where using cloud models costs money, for example Bionic+ ($20/month).
Q. Can I use it at my company? A. Since July 2025, work use has also been free, with no need to contact the company or obtain a commercial license. For organizations that need SSO, model restrictions and similar features, a separate enterprise offering is available.
Q. Can I use it in Japanese? A. The app's interface can be set to Japanese in the settings. You can chat in Japanese by choosing a model that supports Japanese, such as Qwen or Gemma. The website and documentation are in English.
Q. Is what I type sent anywhere? A. The official documentation explains that when you chat with downloaded models or attach documents to a chat, your input and documents do not leave your computer. When you use cloud models in Bionic, they are processed on servers in the US, and the company explicitly states that prompts and responses are not stored.
Q. How do Ollama and LM Studio differ? A. Both run open models on your own machine, but Ollama is command-centric, while LM Studio is an app where you find models, chat and change settings through the interface. LM Studio also has a CLI (lms), and both offer OpenAI-compatible APIs.
Q. Does it work on a Mac? A. Yes, on Apple silicon (M1–M4) Macs running macOS 14.0 or later. Intel-based Macs are not supported.
Q. Will I get a refund if I cancel a Bionic paid plan? A. The Terms of Use state that fees are non-refundable except as required by law or expressly stated otherwise. You can cancel at any time, and cancellation takes effect at the end of the current billing period (as of October 10, 2026).
How LM Studio differs from other AI tools
| Service | Type | Best for |
|---|---|---|
| LM Studio | Desktop app for running local LLMs (GUI + API) | Finding and trying models through an interface, pointing apps at local LLMs |
| Ollama | Local LLM runtime (CLI-centric) + cloud inference | Running models via commands, embedding in servers |
| Hugging Face | Platform for sharing models and datasets | Finding and obtaining a wide range of models, publishing your own |
| Cline | Coding agent that runs in VS Code | Connecting to models in LM Studio and the like to write code |
Qwen and DeepSeek are "models", Hugging Face is the "place" where models are distributed and shared, and LM Studio and Ollama are the "runtime environments" that run those models on your machine.
The information on this page is as of October 10, 2026, and is based on LM Studio's official website (home page, pricing page, download page, LM Link, Enterprise, Terms of Use and official blog) and official documentation (LM Studio, Bionic, developer docs, system requirements and display languages). Pricing and features may change, so please check the official website for the latest information.
