Jan

Jan is a free, open-source, offline-first ChatGPT replacement that lets you run AI models entirely on your device. With over 4 million downloads, Jan provides a beautiful native desktop application for macOS, Windows, and Linux that combines local model inference with optional cloud provider access.

Built by Jan AI, the application supports thousands of models from Hugging Face, including Jan's own optimized Jan-V1 series. It offers a clean, native interface with extensions, multi-model chats, and deep integration for developers. The project is open-source and hosted on GitHub at github.com/janhq/jan.

Jan is available as a native desktop application for all major platforms:

Platform	Download
macOS	Apple Silicon & Intel — download from jan.ai
Windows	Installer for x64 available from the official website
Linux	.deb, .rpm, and .AppImage packages available

Jan provides access to thousands of models through multiple channels:

• Jan-V1 Series — Jan's own optimized model series (Jan-V1-4B, Jan-V1-7B) fine-tuned for the Jan client experience
• Hugging Face Integration — Direct access to any model on Hugging Face with a Hugging Face Pro account
• GGUF Models — Broad support for GGUF format models optimized for local CPU/GPU inference
• MLX Support — Apple Silicon optimized models via MLX for maximum performance on Mac
• Model Explorer — Built-in model browser to discover, download, and switch between models

Jan supports multiple local inference engines to maximize performance on your hardware:

• llama.cpp — CPU-optimized inference with optional GPU acceleration via CUDA, Metal, and Vulkan
• MLX — Apple Silicon-native inference for maximum performance on Mac hardware (up to 90+ tokens/s on Jan-V1-4B)
• TensorRT-LLM — NVIDIA GPU acceleration for Windows and Linux users with compatible GPUs
• ONNX Runtime — Cross-platform inference with optimized performance across different hardware

All local inference runs entirely on-device with zero data leaving your computer.

Jan also supports connection to cloud AI providers for when you need more powerful models or faster responses. You can add multiple providers and switch between them seamlessly:

• OpenAI — GPT-4 and GPT-3.5 models via API key
• Anthropic — Claude models via API key
• Google — Gemini models via API key
• Hugging Face — Use your Hugging Face Pro subscription to access models directly
• Custom Providers — Add any OpenAI-compatible API endpoint

• Offline-First — Full functionality without internet; models run entirely on your device
• Beautiful Native UI — Polished desktop application that feels native on every platform
• Multi-Model Chat — Use different models in the same conversation and compare responses
• Extensions System — Extend Jan's capabilities with community-built extensions
• Local RAG — Chat with your local files and documents without uploading anything
• Privacy-Focused — No telemetry, no data collection, no accounts required for local use

• Official Website — jan.ai
• GitHub — Open-source codebase at github.com/janhq/jan
• Documentation — Guides, API docs, and troubleshooting at jan.ai/docs
• Discord Community — Active community with 15,000+ members
• Changelog — Regular updates and release notes available on the Jan website