Back
LM StudioTRY IT

LM Studio

LM Studio is a desktop app for running AI models locally on your computer. You find and download open models such as Llama, Qwen, Gemma or DeepSeek, chat with them, attach documents, and serve a model through an API for other applications. It's made by Element Labs, and the core app is free for personal and work use.

For a developer, that means testing an AI feature without paying a hosted provider for every request. For anyone handling sensitive documents, it means asking questions about them without the text leaving the machine. LM Studio handles loading and configuring the model. The model you pick decides how good the answers are.

In 2026 LM Studio added Bionic, a separate agent app now featured on its homepage, so the company currently offers two apps. LM Studio is the one for chatting with models, tuning them and running an API. Bionic is for agent work like coding, documents and computer tasks.

Local chat and document questions

You can attach PDF, DOCX and text files to a conversation. Short documents fit directly into the model's context, while longer material uses retrieval to pull out the relevant passages. Ask specific questions and check the answers against the original, because retrieval can miss important information.

Once you've downloaded the models and runtimes you need, local chat, document processing and the local server all work offline. Searching for models, downloading files and checking for updates need an internet connection. External MCP tools and other connected services can send data outside your computer, so running the model locally doesn't make every connected workflow private.

Developer tools

The local API server connects LM Studio to other apps. It offers REST endpoints that are compatible with the OpenAI API, plus Python and TypeScript SDKs, so many tools can point at localhost:1234 instead of a cloud provider. The lms command line tool handles model and server tasks, and llmster is a headless version for servers, cloud machines and CI pipelines that don't need the desktop interface.

MCP support connects models to extra tools. What a model can do with them depends on the model and the server you connect, so review permissions before giving it access to files or external accounts.

LM Link, currently in preview, connects your own machines over an encrypted network built on Tailscale. You can load a model on a powerful desktop or cloud machine and use it from your laptop as if it were local, including from coding tools that talk to the local API. Your chats stay on your device, and only your device list goes to LM Studio's servers.

Bionic, the agent app

LM Studio Bionic with a list of projects on the left, a chat where the agent built a portfolio landing page and lists the edited files, and a live preview of the site on the right

Bionic building a website with a live preview. Image from LM Studio.

Bionic organizes work into projects and sessions, edits documents and code, and can run several sessions at once. It runs local models through LM Studio's runtime, which uses MLX and llama.cpp, or remote models from your own machines through LM Link. For heavier jobs it can switch to large open models such as Kimi, GLM and DeepSeek, hosted by LM Studio in the US with zero data retention. It also includes voice input with transcription that runs on your device.

Hardware and pricing

LM Studio runs on Apple Silicon Macs, Windows and Linux. Intel Macs aren't supported. Memory, GPU, model size and context length decide what runs comfortably, and a large model can exceed a laptop's memory or answer too slowly to be useful.

The LM Studio app is free for personal and work use, though each model has its own license. Bionic's free plan includes the agent, local models, offline voice transcription, LM Link for up to five devices and limited web search. Paid Bionic+ and Pro plans add access to the cloud-hosted open models, more web search and higher usage limits, and extra usage is paid with credits.

Is LM Studio free?

Yes. The LM Studio app is free for personal and work use. Only Bionic's cloud-hosted models and higher usage limits cost money.

How is LM Studio different from Ollama?

Both run open models locally and serve them through an API. LM Studio is built around a desktop app with a model browser, chat and settings for each model, while Ollama started as a command-line tool and is often run in the background for other apps.

Alternatives to LM Studio

  • Ollama: Runs AI models locally on your computer and connects them to coding tools and applications, with optional cloud models.