Jan is a free, open-source desktop app that runs AI models locally on your own hardware or connects to cloud models like GPT, Claude, and Gemini for private chatting.

Follow
Tags
Artificial Intelligence
Last Updated Oct 21, 2026
Jan screenshot

About

Jan is an open-source desktop application that serves as a privacy-focused alternative to hosted chat services such as ChatGPT. It lets users run open large language models directly on their own computers, which means conversations stay on the local machine instead of being processed on someone else's servers. For people who want cloud capabilities anyway, Jan can also connect to online models including OpenAI's GPT, Anthropic's Claude, and Google's Gemini.

The app is developed by Menlo Research and is built openly in public, with its source code available on GitHub, where the project has accumulated roughly 45,000 stars. Jan reports more than 7 million downloads and is free to use. The software ships in two main forms: Jan Desktop, the local chat application, and Jan Agent. A memory feature that carries conversation context between sessions was listed as coming soon at the time of this review.

Key Features

Local model execution: Users can download and run open-source models such as Llama, Mistral, Qwen, DeepSeek, and Gemma on their own hardware, keeping data offline.
Cloud model connections: The app also supports plugging in hosted services like ChatGPT (OpenAI), Claude (Anthropic), and Gemini (Google) for users who want those options.
Open-source development: Jan's code is developed in public on GitHub, so anyone can inspect, audit, or contribute to it.
Cross-model flexibility: The product's model list spans open and proprietary options from Meta, Mistral AI, Alibaba, Google, Moonshot AI, and others, letting users choose per conversation.
Persistent memory (coming soon): A planned feature will let Jan remember context and user preferences across chats so users do not have to repeat themselves.
Desktop app for Mac: The official site offers a direct download for macOS, with version 0.8.7 noted as live on GitHub.
Active community: The project maintains a Discord community and a Hugging Face presence with over a hundred listed models.

Pricing & Fees

  • Pricing: The site describes Jan as free and open source; no paid plans are advertised on the pages reviewed.
  • License and cost: Because it is open source, the software itself carries no purchase price or subscription.
  • Platform availability: A Mac download is offered directly on the homepage; check the official site for other platforms.
  • Notable limitation: The memory feature was marked as "coming soon" and was not yet available at the time of review.

Who It's For

Jan suits developers and technical hobbyists who want to experiment with local large language models without sending data to a cloud provider. It also fits privacy-conscious professionals — lawyers, healthcare workers, or anyone handling sensitive material — who need chat transcripts to remain on their own machines. Students and researchers can use it to compare open models such as Llama, Qwen, and DeepSeek side by side, while casual users who already have API access to GPT or Claude can consolidate those services in one desktop interface.

FAQ

What is Jan?Jan is a free, open-source desktop application that runs AI models locally on a user's computer or connects to cloud models like GPT, Claude, and Gemini, positioned as an alternative to ChatGPT.

Is Jan free?Yes, the official site presents Jan as free and open source, with no paid plans listed on the pages reviewed.

Which AI models can Jan use?Jan supports open models including Llama, Mistral, Qwen, DeepSeek, and Gemma, and can also connect to hosted services from OpenAI, Anthropic, and Google.

Does Jan keep my data private?When running local models, conversations are processed on the user's own hardware rather than a remote server; connecting to cloud models, however, involves those providers' services.

Does Jan have memory?A feature that remembers context and preferences across conversations was listed as coming soon on the official site at the time of this review.