A desktop app for discovering, downloading, and running open models locally on macOS, Windows, and Linux, powered by llama.cpp (GGUF) and Apple MLX. It provides a GUI-first experience with a built-in model search (via Hugging Face), a flexible chat interface, offline document chat (RAG), and a local server exposing OpenAI- and Anthropic-compatible endpoints. 2026 expansion: LM Studio introduced Bionic, a separate AI agent app for coding and document work that can run locally, via LM Link across devices, or against frontier open models on LM Studio Secure Cloud with zero data retention. Local use is free, including at work (commercial restrictions removed July 2025). Cloud plans: Free ($0), **Bionic+** ($20/mo, US-hosted open models, discounted bulk tokens), Pro ($100/mo, 5× usage, early access). Team subscriptions are not yet available; only centralized billing for credits. SDKs (Python/TypeScript), CLI (lms), and a headless daemon (llmster) are also available.

Pros & Cons:

Pros: The GUI-first experience is still the most approachable way to run local models—model search, download, and chat are integrated, and offline document chat works without any cloud dependency; local use is genuinely free, including at work since July 2025, which removes the licensing friction that competing tools sometimes impose; the Bionic expansion means you can start fully local and switch to larger cloud models when your hardware can’t handle the task, with the same zero-data-retention commitment on cloud requests; dual-engine support (llama.cpp and MLX) gives Apple Silicon users native performance, and the OpenAI/Anthropic-compatible endpoints let existing tooling work without changes; SDKs, CLI, and headless mode make it usable beyond the desktop GUI.

Cons: The product line has split—Bionic is a separate app from LM Studio, and understanding which app to use for which task requires reading documentation, which can confuse new users expecting a single download; cloud pricing adds a new cost dimension—Bionic+ at $20/mo includes US-hosted open models, but heavy usage of larger models (e.g., Kimi K3) can exceed the included allowance, and Pro at $100/mo is a significant jump; team features are still incomplete—centralized billing exists, but actual team subscriptions are “coming soon,” so collaborative management is limited compared with Ollama’s Team plan; performance depends on hardware—without a capable GPU (NVIDIA 8GB+ VRAM or Apple Silicon 16GB+), local models run slowly or not at all; the desktop app and Bionic are separate, and settings/configuration do not fully overlap, which adds management overhead for users who want both.

Who is it for?:Users who want a GUI-first way to run open models locally without CLI commands; privacy-conscious individuals and teams that need local inference with offline document chat; developers who want a local server with OpenAI/Anthropic-compatible endpoints for testing and integration; users who want to start local and scale to cloud models for heavier tasks without switching tools entirely; teams evaluating local AI for work (free commercial use since July 2025). Not suited for teams needing full multi-user production serving (vLLM is better), users who want a single unified app for both local and agent work (Bionic is separate), or those seeking team subscription features today (only centralized billing exists).