Jan

Open-source desktop assistant that runs local models and connects to cloud APIs

AI Chat AssistantsFree planOverseas★★★★☆ 4.0

Visit Jan official site ↗ View similar tools Compare alternatives →

What is Jan?

Jan is an open-source desktop application that turns your own computer into the host for an AI assistant. It downloads and runs open-weight models locally through a bundled engine, so a laptop can chat without sending anything to a server, and it can equally be pointed at OpenAI, Anthropic, Gemini or any OpenAI-compatible endpoint. The project aims at a familiar chat experience that stays private and keeps working offline. Its differentiator is the local-first design: model files, conversations and settings live on your disk, and switching between a local model and a hosted one is a dropdown change.

Last updated: 2026-09-20. This site only provides an index; for exact features, pricing, and licensing, see the official website.

Key features

Runs open-weight models locally on CPU or GPU with no separate server
Connects to OpenAI, Anthropic, Gemini and OpenAI-compatible endpoints
Model hub for downloading and swapping local model files
Offline chat with conversations stored on your own disk
Assistant presets and system prompts you can save and reuse
Open-source codebase that also exposes an OpenAI-compatible local API

Pros & cons

Strengths

Works offline and keeps conversations on your own machine
Free and open source with no account required to start
One interface for both local models and hosted providers

Watch out for

Local model quality depends heavily on your available hardware
Fewer built-in agent and tool integrations than larger platforms
Managing model files, sizes and updates takes some patience

Best for & use cases

offline chat, local models, privacy focused work, api experimentation

If you're comparing similar products, check the alternatives below, or browse all tools in the AI Chat Assistants category.

FAQ

Can Jan run without an internet connection?

Yes, once a local model has been downloaded. The app bundles a local inference engine, so chat, model files and history all work offline. Cloud providers are optional additions rather than requirements.

Do I need a GPU?

No, but it helps. Small quantised models run acceptably on a modern CPU, while larger models need a GPU with enough video memory to be pleasant. Speed scales with whatever hardware you have.

Is my data private?

With local models, prompts and replies never leave your machine. If you connect a cloud provider, that provider sees the messages you send it, so the privacy level follows whichever model you select.