Jan is a desktop application for downloading and running large language models locally or connecting to cloud-based models. It provides a graphical interface and a local API server for text-based model interaction without requiring coding skills.
Project overview
The application lets users run LLMs on their own hardware using a desktop GUI while also offering an OpenAI-compatible local API server for other applications.
Project type
MCP · Model Runtime
Use cases
Chat Assistants
Deployment
Refer to project documentation
License
License pending
Best for
General users and developers who want a desktop GUI or a local OpenAI-compatible API to run LLMs without writing code.
Key capabilities
Download and run LLMs (such as Llama, Gemma, Qwen, and GPT-oss) locally from HuggingFace.
Connect to GPT models via OpenAI, Claude models via Anthropic, Mistral, Groq, MiniMax, and others.
Create specialized AI assistants tailored to specific tasks.
Run a local server at localhost:1337 that other applications can connect to.
Use MCP integration for agentic capabilities.
Limitations and risks
Specific RAM requirements scale with model size; for example, macOS requires 8GB for 3B models, 16GB for 7B models, and 32GB for 13B models. Windows requires a GPU for model support.
Connecting to cloud models requires external API credentials for services like OpenAI or Anthropic and may incur paid service costs.
Getting started
Download the installer for your operating system, install the application, and open Jan to begin.
Alternatives and comparisons
Provides a simple way to get up and running with open models locally via CLI or REST API.
Provides LLM inference optimized for Apple Silicon with continuous batching and KV caching managed via a native macOS menu bar app.
Runs Claude Code on-device with local AI on Apple Silicon using an MLX-native Anthropic-API server.