Jan AI Complete Guide — Open-Source Local LLM Desktop App (2026)
Jan is a free, open-source desktop application for running large language models locally on your machine. Licensed under AGPLv3, it gives you a ChatGPT-like interface that runs entirely offline — no data leaves your computer. If you care about privacy, want full control over your AI stack, or need an extensible platform your team can build on, Jan is worth a serious look.
Built by the team at Homebrew Computer Company, Jan is designed around a simple idea: AI should be personal and private by default. The app stores all data — models, conversations, settings — in a local folder you own.
Unlike LM Studio (free but closed-source) or Ollama (CLI-first), Jan combines a polished GUI with full open-source transparency and a plugin architecture.
Key Features
- Offline-first — once you download a model, no internet connection is needed. Your conversations stay on your device.
- Built on llama.cpp — fast inference on consumer hardware with support for GGUF quantized models from HuggingFace.
- Chat interface with threads — organize conversations into threads, switch between models mid-session, and manage chat history locally.
- Local API server — spin up an OpenAI-compatible API endpoint for use with other tools and scripts.
- Remote API fallback — connect to OpenAI, Anthropic, or other providers when you need cloud models alongside local ones.
- Extensions system — add functionality through community and first-party extensions.
- Cross-platform — runs on macOS, Windows, and Linux.
Installation
macOS
Download the .dmg from jan.ai. Open it, drag Jan to Applications, and launch. On Apple Silicon Macs, Jan uses Metal acceleration automatically.
Windows
Download the .exe installer from jan.ai. Run the installer and follow the prompts. NVIDIA GPU users get CUDA acceleration out of the box if drivers are up to date.
Linux
Download the .AppImage or .deb from jan.ai.
For the AppImage:
chmod +x jan-*.AppImage
./jan-*.AppImageFor Debian/Ubuntu:
sudo dpkg -i jan-*.debVulkan support is available for AMD GPUs on Linux.
Downloading and Running a Model
- Open Jan and go to the Hub tab.
- Browse available models or search for a specific one (e.g., Llama 3, Mistral, Phi-3, Qwen 2.5).
- Pick a quantization level —
Q4_K_Mis a good balance of quality and speed for most hardware. If you have 16 GB+ of RAM, you can tryQ5_K_MorQ6_Kfor better output quality. - Click Download. Jan pulls the GGUF file and stores it locally.
- Once downloaded, go to Threads, create a new thread, select your model, and start chatting.
You can also import GGUF models manually. Drop any .gguf file you downloaded from HuggingFace into Jan’s models directory, and it will appear in the model selector. This is useful when you want a specific quantization or fine-tune that isn’t listed in Jan’s built-in hub.
For recommendations on which models to run, see our best AI models for coding locally and best self-hosted AI models guides.
Local API Server
Jan can expose a local API server that’s compatible with the OpenAI API format. This lets you use Jan as a drop-in backend for any tool that supports OpenAI’s API.
- Go to Local API Server in Jan’s settings.
- Select a model and click Start Server.
- The server runs at
http://localhost:1337by default.
Test it with curl:
curl http://localhost:1337/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{
"model": "your-model-id",
"messages": [{"role": "user", "content": "Hello"}]
}'This works with tools like Continue, Open Interpreter, Cursor, or any custom script that targets the OpenAI API. No API key required — just point your tool at localhost:1337 and set any string as the key.
The API supports chat completions, model listing, and streaming responses. If you’re building apps on top of local models, this is one of the easiest ways to get a compatible endpoint running.
Extensions
Jan’s extensions system is what sets it apart from other local LLM tools. Extensions can add:
- New model providers (connect to additional remote APIs)
- Custom tools and integrations
- UI modifications and themes
- Data import/export capabilities
Extensions are managed from the Extensions tab in settings. The architecture is open, so developers can build and distribute their own extensions. This makes Jan particularly appealing for teams that want to customize their local AI setup without forking an entire application.
Jan vs LM Studio vs Ollama
| Feature | Jan | LM Studio | Ollama |
|---|---|---|---|
| License | AGPLv3 (open-source) | Free, closed-source | MIT (open-source) |
| Interface | Desktop GUI | Desktop GUI | CLI-first |
| Model format | GGUF | GGUF | GGUF (via Modelfile) |
| Engine | llama.cpp | llama.cpp | llama.cpp |
| Local API server | Yes (OpenAI-compatible) | Yes (OpenAI-compatible) | Yes (own format + OpenAI-compatible) |
| Extensions/plugins | Yes | No | No |
| Remote API support | Yes (OpenAI, Anthropic) | No | No |
| Model discovery | Good | Excellent | Good (ollama.com library) |
| UX polish | Good | Excellent | Minimal (CLI) |
| Best for | Extensibility, teams, open-source | Ease of use, model exploration | API serving, scripting, Docker |
For a deeper dive, see our Ollama vs LM Studio vs vLLM comparison.
Who Should Use Jan
Privacy-focused users — everything runs locally, the code is auditable, and no telemetry is required.
Teams wanting self-hosted chat — Jan gives you a ChatGPT-like experience you can deploy internally without sending data to third parties.
Developers who want extensibility — the open-source codebase and extensions system mean you can customize Jan to fit your workflow rather than working around limitations.
People who want one app for local and remote models — Jan’s ability to connect to OpenAI or Anthropic alongside local models makes it a unified interface for all your LLM usage.
If you primarily want the smoothest GUI experience and don’t care about open-source, LM Studio might be a better fit. If you need a lightweight API server for production or scripting, Ollama is the stronger choice.
FAQ
Is Jan AI free?
Yes, Jan is completely free and open-source under the AGPLv3 license. There are no subscription fees, usage limits, or hidden costs for running models locally.
How does Jan compare to LM Studio?
Jan is fully open-source and offers an extensions system for customization, while LM Studio provides a more polished GUI with better model management. LM Studio is easier for beginners; Jan is better for users who want transparency and extensibility.
Is Jan open source?
Yes, Jan is fully open-source with its code available on GitHub under the AGPLv3 license. You can audit the code, contribute, and verify that no telemetry or data collection is happening.
Does Jan work offline?
Yes, Jan works completely offline once you’ve downloaded your models. All inference runs locally on your hardware with no internet connection required for chat or API server functionality.