Jan

Developed by the janhq team and available on GitHub with full source code transparency, Jan puts the user in complete control of their AI interactions all processing happens locally, no data is sent to external servers, and conversations are never used to train third-party models.

It runs on Windows 10/11, macOS (both Intel and Apple Silicon), and Linux distributions including Ubuntu, Debian, and Fedora.

Jan's integrated Hub makes discovering and downloading AI models straightforward for non-technical users. The Hub provides a curated browser for models from sources like Hugging Face, with hardware guidance indicators that warn users if a model may be slow on their device or requires more RAM than they have available.

Supported model families include Llama, Gemma, Qwen, and many others, giving users access to the full breadth of open-source language models available in the research and developer community without manual configuration.

Beyond running models locally, Jan also supports connecting to cloud-based AI providers when an internet connection and API access are available.

Users can configure integrations with OpenAI (GPT models), Anthropic (Claude), Mistral, Groq, MiniMax, and other services through the same unified interface, making Jan a versatile multi-model chat frontend as well as a local inference platform.

This hybrid capability allows users to switch between local privacy-first models and cloud models depending on their needs and connectivity.