LM Studio downloads and runs open language models on your own computer, with a catalogue for finding them and a desktop application for using them. Its Bionic agent works on documents, slides and software against either local models or frontier ones.
Local execution is the point. Nothing leaves the machine, which changes what you can reasonably put through a model: contracts, patient notes, unreleased code and anything else where sending text to someone else’s server is the objection rather than the cost.
The application is free and Bionic carries per-token usage pricing above a free tier. Local inference needs capable hardware, and open models still lag the frontier ones on genuinely hard problems, which is the trade you accept for privacy.
