LM Studio launches Bionic, an AI agent built for open models

LM Studio has announced Bionic, a separate application it describes as an AI agent built for open models, aimed at coding, research and work with documents and files. The company says users can run models locally or switch to open-source models in the cloud for heavier tasks while keeping control of their privacy and spending, and commits to zero data retention and never training on user data for all Bionic users.

Voice input that never leaves the device

Bionic ships with a voice keyboard that transcribes speech locally, using on-device audio models. For launch LM Studio is bundling Mistral AI's Voxtral, a multilingual real-time transcription model. The keyboard can be started from any application and Bionic will transcribe at the cursor, so dictation works across the system without sending audio away.

Coding, documents and sandboxed work

A Code project can be pointed at a local folder, after which the agent inspects the codebase, explains unfamiliar code, makes edits and debugs, with inline diffs for reviewing each change and agentic search for tracing behaviour. LM Studio names GLM 5.2 and Kimi K2.7 Code as models the agent works with. In a Work project, Bionic processes documents, PDFs, decks and spreadsheets inside a sandbox, organises directories, summarises material and pulls in outside context through native web search; automatic checkpoints allow changes to be reviewed or rolled back, and in-app previews keep the material in one place.

Cloud inference and cost control

Models can run locally on the LM Studio runtime, be reached through LM Link, or be used from the LM Studio Secure Cloud, which hosts frontier open models. Cloud requests are processed transiently and are not retained after the request completes, the company says. Choosing the right model and compute for each task is presented as the main lever for controlling cost.

Bionic is a new app that sits alongside LM Studio, which remains available for advanced low-level configuration. Using cloud models requires an LM Studio account with billing set up; from there a user connects a project, picks a model and starts working.