Running models on your own machine
LM Studio is a desktop application for Windows, Mac, and Linux that downloads published AI models to your computer and runs them locally. Finding a model, loading it, and chatting with it all happen in the interface.
Rather than sending questions to a cloud service, it runs supported models on your own CPU or GPU. Once a model is downloaded, chat, document reading, and a local API server all work offline.
What it suits
- Trying published models such as Qwen, Llama, and Mistral
- Having AI read text without sending it to an outside server
- Comparing the effect of model size and quantization
- Calling a local API from your own application
- Chatting where there is no internet connection
Because the basics stay in the interface, it is an approachable first step into local LLMs. Searching for and downloading models, and updating the app, still need a connection.
How it differs from Ollama
Ollama also runs models locally. LM Studio makes browsing models and chatting easy in one window; Ollama is cleaner to drive from a terminal or an API.
Both can be connected to from other AI applications. LM Studio if you want to manage models visually, Ollama if you want to build it into scripts or a development setup — though using both is perfectly reasonable.
Choosing a model
A bigger model does not necessarily run comfortably on your machine. Check memory requirements, supported formats, quantization, and context length, and start small. Watch heat and free disk space too.
Running locally does not remove licensing. Check the terms for commercial use and redistribution per model. And exposing the local API to a network invites outside use, so set the binding and authentication appropriately.