LM Studio is a free desktop app for running open-source language models on your personal computer. Unlike command-line tools, everything happens in a familiar graphical interface: you search for a model, pick the version that fits your hardware, download it and chat with it in a window that looks like ChatGPT. You can also give it files to answer from, and with one switch start a local OpenAI-compatible server so other programs can connect to it. Its main difference from Ollama is that full interface and the guidance on which model will run on your system.

What is LM Studio good for?#

  • Private chatbot: someone who doesn't want personal notes and text going to an outside server chats with a local model.
  • Working offline: during an outage or a slow connection, a downloaded model keeps working.
  • Asking questions of a document: give it a PDF or text file and ask questions about it without sending the file anywhere.
  • Comparing models: before choosing a model for a project, try several on the same question.
  • Coding backup: connect the local server to your code editor extension or your own Python script.

Strengths and weaknesses#

Strengths:

  • A simple graphical interface, suitable for someone who has never used a command line.
  • Shows which model versions fit your system's memory.
  • Good support for Apple-silicon Macs and common graphics cards.
  • A local OpenAI-compatible server.

Weaknesses:

  • The app itself is not open source.
  • Heavier than Ollama and not the first choice for a headless server.
  • The main limit is the one every local model has: lower quality than large cloud chatbots, and the need for strong hardware to run big models.

Free tier and pricing#

LM Studio is free for personal and work use and needs no separate license. An enterprise edition with features such as single sign-on and model controls is also available for companies. See the official site for current details.

Language support#

The app's interface is in English, and quality in other languages depends on the model you choose. Small models are weak outside English, while mid-sized models from families such as Gemma and Qwen give more acceptable results. If you have a model file from another source, you can place it manually in the models folder.

Getting started in three steps#

  1. Install the app from the official site.
  2. In the search section, pick and download a small recommended model.
  3. Open a new chat, load the model and, if needed, turn on the local server.

LM Studio alternatives#

  • Ollama: lighter, open source and more popular with developers and other tools.
  • Hugging Face: the main model repository, for finding specialized and multilingual models.
  • Gemma: a family of light models that run well on an ordinary machine.

To choose hardware and a model, read running AI on a personal computer and open-source AI models: what they are and when they help.