Jan

What is Jan?

Jan is a free, open-source desktop application for running AI models locally without any internet connection, cloud dependency, or account requirement. Built as a privacy-first alternative to ChatGPT and other cloud AI services, Jan lets you download and run language models like Llama 3, Gemma 3, Mistral, Qwen, and any GGUF-format model directly on your own hardware. Every conversation stays on your device — no data is ever sent to external servers.

Jan stands out from other local AI tools like Ollama and LM Studio with its polished desktop interface and focus on user experience. While Ollama is command-line-first and LM Studio targets developers, Jan provides a ChatGPT-like conversation interface that non-technical users can navigate comfortably. The built-in model hub lets you browse, download, and manage models with one click, without touching terminal commands or configuration files.

Under the hood, Jan includes an OpenAI-compatible local API server, a VSCode-like extension system for adding custom functionality, and support for both CPU and GPU inference. For users who want cloud capabilities alongside local models, Jan also supports connecting to remote APIs like OpenAI, Anthropic, and Groq via API keys. The in-house Jan-v2 model adds multimodal capabilities, and Jan-v1 includes web search functionality.

Jan is completely free and open-source under the Apache 2.0 license. There are no pricing tiers, subscriptions, or usage limits. The only cost is your own hardware — Jan runs on macOS, Windows, and Linux. A 16GB RAM machine with a modern CPU can run 7B parameter models comfortably; for larger models (13B-70B), a dedicated GPU is recommended.

Jan Pricing

Completely free and open-source

Key Features

• 100% offline capability — Run AI models completely locally with no internet connection, cloud dependency, or data leaving your device
• ChatGPT-like desktop UI — Polished conversation interface with chat history, model switching, and settings that non-technical users can navigate easily
• One-click model downloads — Built-in model hub to browse, download, and manage models like Llama 3, Gemma 3, Mistral, and Qwen without terminal commands
• GGUF model support — Import any GGUF-format model from Hugging Face or other sources for maximum flexibility in model selection
• OpenAI-compatible API server — Local API server that mimics the OpenAI API, allowing existing tools and scripts to connect to your local models
• Extension system — VSCode-like plugin architecture for adding custom functionality, integrations, and workflows
• Cloud API connections — Optionally connect to OpenAI, Anthropic, Groq, and other cloud providers via API keys alongside local models
• CPU and GPU inference — Automatic hardware detection with optimized inference for both CPU-only and GPU-accelerated setups
• Cross-platform support — Available on macOS (Apple Silicon and Intel), Windows, and Linux with consistent experience across platforms
• In-house models — Jan-v2 with multimodal capabilities and Jan-v1 with web search functionality built specifically for the platform
• OpenAI-compatible API
• Extensions system
• GPU acceleration
• Conversation history
• Multiple model support

Pros & Cons

Best For

Privacy-focused individuals who want AI assistance without any data leaving their device or any account creation. Developers who need a local OpenAI-compatible API server for testing and development without API costs. Non-technical users who want to try local AI models without learning command-line tools or Docker. Organizations with strict data policies that prohibit sending confidential information to cloud AI services.