← All tools

Tool

LM Studio

Local LLMs · Desktop app · lmstudio.ai

A desktop app for discovering, downloading, and running open-weight language models on your own computer — with a chat UI and an OpenAI-compatible local server when you need an API.

What it is

LM Studio (from Element Labs, Inc.) is a GUI-first local LLM runner for Windows, macOS, and Linux. You browse models in-app, download weights, chat locally, and optionally expose a local API so other tools can talk to the same model. The company also ships developer SDKs and optional paid/cloud features (for example agent and cloud inference offerings branded around “Bionic”); those are separate from the core “run a model on this machine” path.

What it replaces

For interactive local chat and for wiring other apps to a local OpenAI-compatible endpoint, it replaces “pay a hosted chat subscription for every draft” when open weights are good enough and you want the prompts to stay on-device. It also replaces rolling a raw llama.cpp / MLX setup if you want a polished desktop shell.

It does not replace a frontier hosted API when you need the strongest closed models, nor does it replace a production inference cluster for a team. Optional cloud/agent features are not the same product as offline local inference — check current pricing and data-handling docs if you leave the fully local path.

License

The LM Studio desktop app is proprietary software, not an open-source project. Per the published App Terms of Service (version noted on lmstudio.ai/terms as of August 23, 2026), Element Labs grants a non-exclusive, non-transferable license to use the software for personal and/or internal business purposes. The company’s own materials state that LM Studio is free for home and work use under those terms; optional paid features and an Enterprise offering exist for orgs that need more than the free desktop path.

Model licenses are separate. Weights you download (Llama, Gemma, Qwen, etc.) carry their own terms. The app license does not rewrite a model’s license. Always read the model card before commercial use or redistribution.

This is a summary of public terms, not legal advice. Re-read lmstudio.ai/terms before you rely on it in a compliance review — terms can change.

Where it falls over

  • Hardware still wins. Local quality and speed track your GPU/CPU, RAM, and disk — not the logo on the window.
  • Proprietary app: you get a free-to-use binary under their terms, not “fork and rebrand the product.”
  • Model library ≠ endorsement. A downloadable weight can still have use restrictions.
  • Exposing the local server on a network without thinking is still an unauthenticated model endpoint waiting to happen.
  • Cloud/agent add-ons (when you use them) are a different trust and billing boundary than offline local chat — verify ZDR/privacy claims on the current docs, don’t assume.

Who it's for

People who want a desktop-native local LLM experience with less CLI ceremony than Ollama-only workflows, and who are fine living under a proprietary EULA in exchange for that polish. Good fit for private drafting and local API experiments. Not a magic frontier-model substitute.

Hardware that usually matters

The apps on these pages are free or free-to-use. The recurring cost is usually the machine: VRAM first, then system RAM, then disk for weights. Example Amazon listings skewed toward entry and mid budgets for beginners checking fit — not endorsements; stock and price move constantly.

Some links are Amazon Associates. We may earn a commission if you buy.

Verdict

Use LM Studio when you want a local GUI + local API and the open-weight bar is enough. Treat the app as free-for-personal-and-internal-business proprietary software under current published terms, and treat every model’s license as its own homework. Keep a hosted frontier model for the hard cases.

← Tools · Local LLMs · Home · lmstudio.ai