Jan is an open-source desktop app from Menlo Research that aims to replace ChatGPT with something running on your own computer. You download open models, chat with them, and your data never leaves the device. If needed, you can also connect to cloud services with your own API key.

What is Jan good for?#

  • Fully offline chat: write sensitive notes and drafts with a model without sending anything to the internet.
  • Trying open models: students and enthusiasts test different models from a built-in library.
  • A local server for developers: Jan can offer an OpenAI-compatible local API so your apps can connect to it.
  • One hub for models: manage local models and cloud models with your own key in a single interface.
  • A cheap option for simple tasks: summarizing, rewriting and light Q&A without a monthly subscription.

Strengths and weaknesses#

Strengths:

  • Free, open source and needs no account.
  • Good privacy, since processing is local.
  • A simple interface suitable for beginners.
  • Can connect to cloud models and offer a local API.

Weaknesses:

  • Answer quality is limited by the model and your hardware.
  • Slow on a weak computer.
  • Fewer features than the big cloud tools.
  • Support for many languages is weaker than in cloud chatbots.

Free version and pricing#

Jan is completely free and open source, with no subscription. The only costs are your own hardware and, if you use them, the API charges of cloud models. See the official site for details.

Language support#

The app interface is mainly in English, and quality in other languages depends on the chosen model. Local installation is usually simple, but downloading the app and models from external repositories may be difficult in some regions.

Getting started in three steps#

  1. Install the app for your operating system.
  2. Download a model from the built-in library that fits your hardware.
  3. Start chatting; add a cloud API key later if you need it.

Jan alternatives#

  • LM Studio: a desktop app with more control over models and settings.
  • GPT4All: another lightweight app for local chat and talking to documents.
  • Ollama: a command-line engine for running models, good if you're comfortable with a terminal.
  • Open WebUI: a self-hosted, multi-user web interface for local models.