LM Studio
LM Studio enables running local AI models such as gpt-oss, Llama, Gemma, Qwen, and DeepSeek privately on your computer. It provides a GUI application for desktop use and a headless version, llmster, for server deployments on Linux, cloud, or CI environments. Developers can leverage JS and Python SDKs and an OpenAI compatibility API for integration.
Visit LM StudioExternal link — opens lmstudio.ai in a new tab. LM Studio is a third-party product; we are not affiliated with it.
Link checked 19 September 2026: this site responded and still names the product.
About LM Studio
What it is
LM Studio is a desktop application for running language models on your own computer. It provides a chat interface, downloads models for you, and can serve them locally to other software through an API that mimics the common one.
Why it's different
It makes local models genuinely approachable, which was the real barrier rather than the models themselves, since the alternative involves a command line, quantisation formats and dependency problems. Everything stays on your machine, so it works offline and with material that cannot be uploaded anywhere. The honest limits are hardware and quality: you need a reasonably powerful computer with plenty of memory, and a model small enough to run locally is meaningfully weaker than a frontier one.
How people use it
Working with confidential material that cannot leave the machine. Using a model offline. Experimenting with open models without any setup. Serving a local model to other applications. Try the largest model your memory will hold rather than the smallest that runs, because the gap between local and hosted narrows considerably at the top of what your machine manages.
Step by step
- Download the model first, then chat — the search tab shows file sizes, and that number is roughly the memory it needs.
- Pick a quantised version rather than the full-precision one. It is the difference between a model that fits and one that does not.
- Raise the GPU offload slider until it stops being stable, then step back one. That single setting decides whether responses are fast or unusable.
- Start the local server tab if other software should use it, which gives you an endpoint most tools accept in place of a cloud provider.
- Try the largest model your memory holds before judging local AI. Most disappointment comes from testing a tiny one picked for speed.
Written by the n3os team. We are not affiliated with LM Studio.
This listing was written from public information, without LM Studio’s involvement. If you own it and something here is wrong — or you would rather not be listed at all — email us and we will correct or remove it.
Get the ones worth knowing about
We write one of these for every tool worth the trouble. Get the new ones, plus what we have found genuinely useful lately.
Your address goes to Buttondown, who send the emails on our behalf. One click unsubscribes, and the list is never sold or shared.