Jan is a free, open source ChatGPT alternative that runs AI models locally and offline. Compare its features, model support, and alternatives like LM Studio.
Jan is an open source desktop application that lets anyone run AI chat models locally on their own computer, offering a ChatGPT-style interface without requiring cloud access or an API subscription. It is built by Menlo Research, a Singapore company founded in 2023 by Stanford alumni Daniel Ong and Nicole Zhu, with the project quickly growing to over 42,000 GitHub stars and 5 million downloads.
Under the hood, Jan is a Tauri and TypeScript desktop app powered by the llama.cpp inference engine, which lets it run quantized open-weight models efficiently on ordinary consumer hardware. It ships under the Apache 2.0 license, meaning the full source is auditable and free to adapt, including for commercial use.
Jan includes an in-app model catalog for downloading open-weight models like Llama, Qwen, Gemma, Mistral, DeepSeek, and GPT-OSS, with over 120 models accessible via Hugging Face integration, and it works entirely offline with no telemetry by default. Users can also connect it to cloud APIs such as OpenAI, Anthropic Claude, or Google Gemini using their own API keys, mixing local and cloud inference inside one chat interface.
For developers, Jan can run as a local server that mimics the OpenAI API format, so existing applications built against OpenAI's API can point at Jan and use local models as a drop-in substitute with no code rewrite, which has made it popular for prototyping and privacy-sensitive development workflows.
Jan is completely free and open source, with no subscription tiers, seat-based pricing, or feature paywalls; every capability, including the model catalog, local API server, and cloud model integrations, is available at no cost.
The only costs a user might incur are indirect: sufficiently powerful hardware to run larger local models comfortably, or usage fees charged by third-party providers like OpenAI or Anthropic if the user chooses to connect their own API key for cloud models within Jan.
Yes, Jan is completely free and open source under the Apache 2.0 license, with no subscription fees.
No. Jan can run entirely offline using locally downloaded open-weight models; an internet connection is only needed if you choose to connect a cloud model API.
Jan is built by Menlo Research, a Singapore-based company founded in 2023 by Daniel Ong and Nicole Zhu.
Jan supports open-weight models including Llama, Qwen, Gemma, Mistral, and DeepSeek locally, plus cloud models like ChatGPT, Claude, and Gemini via API key.
Yes, Jan can run as a local server exposing an OpenAI-compatible API, letting developers point existing applications at local models with no code changes.
Jan is available as a native desktop application for macOS, Windows, and Linux.