LM Studio
Free local LLM app, now with the Bionic agent and paid cloud tiers
Overview
LM Studio, from New York's Element Labs, is a desktop app for downloading and running open-weight LLMs locally through llama.cpp and MLX, with an OpenAI-compatible server. It is free for home and work use. Its Bionic agent adds optional paid US-hosted cloud models.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 12 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- Bionic Agent
- Run local LLMs using llama.cpp and MLX
- Offline voice transcription
- LM Link for up to 5 devices
- Limited web search
- US-hosted open models such as Kimi K3, GLM 5.3 and DeepSeek V4 Flash
- Discounted bulk tokens
- Web search and page extraction
- 5x usage limits
- Discounted bulk tokens
- Early access to new features
- US-based zero-data-retention cloud, your own infrastructure or edge devices
Use cases
What you can produce with LM Studio
- Open-weight models browsed and downloaded from Hugging Face inside the app, with hardware-fit guidance
- Offline chat with local models, with document attachments for private RAG
- A local OpenAI- and Anthropic-compatible API server, plus the headless llmster daemon for servers and CI
- One model split across several NVIDIA GPUs with CUDA tensor parallelism
- The MLX engine on Apple Silicon, with KV-cache checkpointing for long agent sessions
- Models on your other machines through LM Link, including from the Locally iPhone and iPad app
- The Bionic agent for coding and document work, with optional paid US-hosted cloud models
LM Studio remains one of the most approachable ways to run open-weight LLMs on your own machine: a Hugging Face model browser, offline chat with document attachments and a local OpenAI-compatible server, all free for home and work since July 2025. It has also grown beyond the desktop, with a headless daemon for servers, multi-GPU tensor parallelism, LM Link between your own devices and an iPhone client.
The July 2026 Bionic agent adds paid cloud tiers at $20 and $100 a month on US-hosted servers with zero data retention, which teams with data-residency rules should weigh. The app itself is proprietary, though its CLI and SDKs are MIT-licensed.
What we found
LM Studio is a desktop app for downloading and running open-weight language models on your own computer, made by Element Labs, Inc. in New York. It launched in May 2023 and runs GGUF models through llama.cpp on Windows, macOS and Linux, plus Apple's MLX engine on Apple Silicon Macs. The model browser pulls from Hugging Face, and a local server exposes OpenAI-compatible and Anthropic-compatible endpoints.
Two changes matter for anyone who last looked a year ago. In July 2025 Element Labs dropped the rule that work use needed a separate commercial licence: the app is now free "both at home and at work". And in July 2026 it launched LM Studio Bionic, an agent app for coding and document work that can run local models or switch to open models hosted in LM Studio's US cloud. That cloud is where the money now comes from: Bionic+ costs $20 a month and Pro $100 a month, while the local app and the free Bionic tier cost nothing. Enterprise deals go through a contact form. Earlier versions of this page called it free-only and desktop-only.
The local runtime has kept improving. Version 0.4.0, in January 2026, introduced llmster, a headless daemon for servers and CI, and parallel requests with continuous batching. Version 0.4.15 added CUDA tensor parallelism for loading one model across several GPUs, and a June 2026 MLX engine update added disk-backed KV-cache checkpointing for long agent sessions. LM Link connects your own machines over an end-to-end encrypted link, and after buying the Locally AI apps in April 2026, Element Labs relaunched Locally as its iPhone and iPad client.
The desktop app itself is proprietary, though its command-line tool, its JavaScript and Python SDKs and the MLX engine are MIT-licensed on GitHub. Bionic's cloud models run on US-hosted servers under a zero-data-retention policy, which teams with residency rules should check against their own requirements. For private local inference behind a friendly interface, it remains a straightforward place to start.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including LM Studio's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with LM Studio unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to LM Studio directly →
Spotted something out of date? Suggest an update →
More in LLMs & Chat
