LM Studio launched Bionic on July 16, a new standalone application built as an AI agent for open models, separate from its existing local LLM runtime and targeting real-work tasks including coding, research, and document-heavy projects.
LM Studio says Bionic works with capable open models including GLM 5.2 and Kimi K2.7 Code across three execution modes: running models locally, connecting through LM Link (the company's remote compute relay), or routing heavier tasks to LM Studio Secure Cloud, which processes requests transiently with Zero Data Retention by default. That privacy-first positioning is a deliberate response to cloud AI services that train on user inputs, which matters most for developers and researchers handling proprietary codebases or sensitive enterprise documents.
Bionic's coding mode can inspect local codebases, trace unfamiliar code, and apply edits with inline diffs so every change is reviewable before it lands. A separate Work project mode sandboxes document operations, letting Bionic process PDFs, slides, and spreadsheets in an environment isolated from the rest of the filesystem. Automatic checkpoints let users review or roll back any changes, and native web search can pull in outside context mid-session.
Voice input ships at launch through a local transcription layer built on Voxtral, Mistral AI's multilingual realtime transcription model. The voice keyboard transcribes directly into whichever app has cursor focus, fully on-device with no audio sent to external servers. That distinction is meaningful for developers on regulated teams or in environments where audio logging is a compliance concern.
Bionic is a strategic pivot for LM Studio, which has spent the past two years building a developer-focused runtime for locally running open-weight LLMs. Agentic coding tools like Cursor, Windsurf, and GitHub Copilot dominate the IDE-integrated market today, but each relies on proprietary cloud inference. Bionic's pitch is the same agentic reach with open-weight models and user-owned data. The decision to also offer cloud-hosted frontier open-source models through LM Studio Secure Cloud suggests the team knows local-only execution still trails closed-model performance on demanding coding tasks, and wants a competitive answer while local hardware catches up.
Bionic downloads as a separate app from the existing LM Studio client, which remains available for low-level model configuration and developer workflows. Cloud inference requires an LM Studio account with billing enabled. The company behind both products is Element Labs, Inc.
