jan

Jan is an open-source desktop application positioned as a local alternative to ChatGPT. According to its README, it lets users download and run large language models directly on their own computer, with an emphasis on full control and privacy, and it can operate entirely offline. Key capabilities described in the README: - Local AI models: download and run LLMs such as Llama, Gemma, Qwen and GPT-oss from HuggingFace. - Cloud integration: optionally connect to GPT models via OpenAI, Claude models via Anthropic, plus Mistral, Groq, MiniMax and others. - Custom assistants: create specialized AI assistants for particular tasks. - OpenAI-compatible API: a local server exposed at localhost:1337 so other applications can use it. - Model Context Protocol (MCP) integration for agentic capabilities. - Privacy-first design: everything runs locally when the user wants it to. Installation is offered through prebuilt downloads for Windows (jan.exe), macOS (jan.dmg), and Linux (deb, AppImage, plus an Arm64 how-to), as well as Microsoft Store and Flathub listings. The project can also be built from source; prerequisites include Node.js 20+, Yarn 4.5.3+, Make, and Rust for Tauri, with make targets for dev, build, test and clean. Windows builds are documented as requiring Git Bash, Visual Studio 2022 Build Tools, LLVM, Ninja and CMake, with optional CUDA Toolkit for CUDA engine variants selected via JAN_ENGINE_VARIANT. Stated system requirements mention macOS 13.6+ with RAM guidance by model size (8GB for 3B, 16GB for 7B, 32GB for 13B), Windows 10+ with GPU support for NVIDIA/AMD/Intel Arc, and most Linux distributions with optional GPU acceleration. The project is licensed under Apache 2.0 and credits Llama.cpp, Tauri and Scalar as foundations. Documentation, API reference, changelog, Discord community and GitHub issue tracker are linked for support and contribution.