LM Studio
VerifiedDesktop app for running local LLMs, now with the Bionic agent and hosted open models
Product preview
LM Studio
What it is
LM Studio is a desktop application for running large language models on your own machine. It bundles a model browser, a chat interface and a local runtime built on llama.cpp and Apple's MLX, so you can download an open model and start a private conversation without touching the command line. The current app centers on Bionic, an agent that can create and edit documents, help with code, run automations and transcribe voice locally. Developers get a CLI, a headless server called llmster and a programmable API. The free plan covers local models, offline transcription, limited web search and LM Link across up to five devices, while paid plans add US-hosted open models such as large Kimi, GLM and DeepSeek releases.
Key features
- Model browser with one-click local downloads
- Bionic agent for documents, code and automations
- llama.cpp and MLX runtimes for fast local inference
- CLI, headless server and developer API
What we liked
- + Friendly graphical setup for local models
- + Strong performance on Apple Silicon through MLX
- + Free tier covers full local use
Worth noting
- · Large models need a lot of RAM or VRAM
- · Hosted models and higher limits require a subscription
Editor's notes
More context from AI BirdView
LM Studio is a desktop app that lets anyone run open language models locally on Mac, Windows or Linux through a graphical interface instead of a terminal. You search a model catalog, download a version sized for your hardware, and chat with it privately. Its newer angle is Bionic, an agent layer that turns the app from a simple local chat window into a workspace that can draft and edit documents, work on code and take actions, while still defaulting to models on your own machine.
What is LM Studio used for?
Most people use LM Studio to try open models privately, to keep confidential work off cloud chatbots, and to give their own scripts a local model endpoint.
- Browsing and downloading open models, with runtimes based on llama.cpp and MLX
- Chatting with local models through a clean desktop interface
- Using the Bionic agent to create and edit documents that save automatically
- Transcribing voice locally in several languages
- Serving models to other apps through the developer API, the lms CLI or the llmster headless daemon
- Linking several of your devices with LM Link so one machine can serve models to another
Who LM Studio is best for
LM Studio suits people who want local AI without learning command-line tools: writers handling sensitive drafts, consultants with client data, students on capable laptops, and developers who want a GUI for testing models before wiring them into code. Apple Silicon owners benefit from the MLX engine. If your computer has limited memory, or you want frontier closed models like the latest GPT or Claude, a hosted assistant will serve you better.
LM Studio pricing
The Free plan costs $0 and includes the Bionic agent, local models through llama.cpp and MLX, offline voice transcription, LM Link for up to five devices and limited web search. Bionic+ is $20 per month and adds US-hosted open models such as Kimi K3, GLM 5.3 and DeepSeek V4 Flash, discounted bulk tokens and web search with page extraction. Pro is $100 per month with five times the usage limits and early access to new features. LM Studio says team subscription plans are coming soon, with centralized billing for inference credits available through organizations now. Prices reflect the pricing page as of October 2026.
Limitations to keep in mind
- Model quality and speed depend on your hardware; small machines are limited to small models
- Hosted models move prompts off your device, so check which mode you are using for sensitive work
- Team plans were still marked as coming soon in October 2026
- The pricing page does not spell out commercial licensing terms, so businesses should read the terms of use
LM Studio vs alternatives
ChatGPT and Claude are better choices when you need the strongest closed models, rich integrations and no hardware constraints. Together Chat lets you try large open models in a browser without downloads. Vibe (Mistral) offers a hosted assistant built on Mistral's models with a European focus. LM Studio wins when privacy and offline use matter most and you want a polished app rather than a terminal tool.
Tips for getting started with LM Studio
Pick a quantized model that leaves headroom in memory, since running at the edge of your RAM slows everything down. On a Mac, prefer MLX builds of a model when they are offered. Turn on the local server only when another app needs it, and use LM Link if you have a powerful desktop and a lighter laptop.
LM Studio FAQ
- Is LM Studio free?
- Yes. LM Studio's Free plan covers local models, the Bionic agent and offline transcription at no cost. Paid plans start at $20 per month for hosted open models.
- Does LM Studio work offline?
- Yes. Once you download a model, LM Studio can run it and transcribe voice entirely on your computer without an internet connection.
- Does LM Studio have an API?
- Yes. LM Studio offers a programmable API, the lms command-line tool and a headless server called llmster for servers and CI.
- What is the best alternative to LM Studio?
- For hosted frontier models, ChatGPT or Claude are the usual choices. For open models in the browser without a download, Together Chat is a simple option.
Reader reviews
From people who've used LM Studio. Emails are never shown publicly.
No reviews yet. Be the first to share your experience.
Alternatives to LM Studio
Other tools we like in this space.
ChatGPT
The assistant that started it all
Claude
Anthropic's thoughtful assistant for serious work
Vibe (Mistral)
Formerly Le Chat: Mistral's fast, privacy-minded assistant with Chat, Work, and Code modes
Together Chat
Free chat app from Together AI running open-source models hosted in North America
Ollama
Run open AI models on your own computer, or call larger ones on Ollama's cloud
Jan
Free, open-source ChatGPT alternative that runs local models or connects to cloud APIs