Run a Private ChatGPT on Your Computer with Jan

// Before you start

What you need

A computer running Windows 10+, macOS 13.6+, or Linux
At least 8 GB of RAM (16 GB recommended for better models)
About 5–10 GB of free disk space for the app and a model
No coding skills needed — this is a regular desktop app install

Step 1 of 5

Download the Jan installer for your operating system

2 min

Jan is distributed as a normal desktop installer — just like installing any other app. Go to the official releases page and grab the file that matches your system. Windows users want the .exe, Mac users want the .dmg, and Linux users can choose .deb or .AppImage.

Terminal · mac

$ Open your browser and go to: https://github.com/janhq/jan/releases/latest

$ Download the correct file:

$ Windows → jan-win-x64-<version>.exe

$ macOS → jan-mac-<version>.dmg

$ Linux → jan-linux-amd64-<version>.deb OR jan-linux-x86_64-<version>.AppImage

What you should see

A single installer file appears in your Downloads folder.

This might happen

macOS may say the app is from an unidentified developer and block it.

Open System Settings → Privacy & Security, scroll down, and click 'Open Anyway' next to the Jan entry. This is a one-time step.

Step 2 of 5

Install and launch Jan

3 min

Run the installer the same way you install any desktop app. On Windows double-click the .exe and follow the prompts. On macOS open the .dmg and drag Jan to your Applications folder. On Linux either double-click the .deb to install via your software center, or make the AppImage executable and run it. Jan will open to a welcome screen when done.

Terminal · mac

$ Windows: Double-click jan-win-x64-<version>.exe and click through the installer wizard.

$ macOS: Open jan-mac-<version>.dmg → drag Jan to Applications → open Jan from Applications.

$ Linux (deb): sudo dpkg -i jan-linux-amd64-<version>.deb

$ Linux (AppImage): chmod +x jan-linux-x86_64-<version>.AppImage && ./jan-linux-x86_64-<version>.AppImage

What you should see

The Jan desktop window opens showing a welcome or home screen.

This might happen

Linux .deb install may report missing dependencies.

Run: sudo apt-get install -f This automatically pulls in anything missing, then try launching Jan again.

Step 3 of 5

Download your first AI model inside Jan

10–30 min (depends on your internet speed)

Jan has a built-in model hub. You browse and download models directly inside the app — no command line needed. Smaller models (3B parameters) run fine on 8 GB RAM. If you have 16 GB, try a 7B model for noticeably better answers. The download happens once; after that everything runs offline.

Terminal · mac

$ Inside Jan:

$ 1. Click the 'Hub' or 'Explore Models' icon in the left sidebar.

$ 2. Browse the list — look for a model tagged '3B' or '7B'.

$ Good starter picks: Llama 3.2 3B, Qwen 2.5 3B, or Gemma 3 4B.

$ 3. Click the model name, then click the 'Download' button.

$ 4. Wait for the progress bar to reach 100%.

What you should see

The model shows a green 'Downloaded' or 'Use' badge. Download size is typically 2–5 GB.

This might happen

Download stalls or fails partway through.

Check your internet connection, then click the model again and resume or restart the download. Jan supports resuming interrupted downloads.

Step 4 of 5

Start a chat and test the model

2 min

Once a model is downloaded you can start chatting immediately. Jan loads the model into your RAM and runs it entirely on your machine. The first response after loading may take 10–30 seconds; after that replies come steadily. Nothing you type leaves your computer.

Terminal · mac

$ Inside Jan:

$ 1. Click 'New Thread' or the chat icon in the left sidebar.

$ 2. Make sure your downloaded model is selected in the model picker at the top of the chat window.

$ 3. Type a message in the text box at the bottom and press Enter.

What you should see

Jan shows a loading spinner briefly, then the model starts streaming a reply word by word in the chat window.

This might happen

The app freezes or is very slow when loading the model.

Your RAM may be too full. Close other apps (especially browsers with many tabs), then try again. If you have only 8 GB RAM, stick to 3B models only.

Step 5 of 5

(Optional) Connect a cloud AI like ChatGPT or Claude

3 min

Jan can also act as a single interface for cloud AI services. If you have an OpenAI, Anthropic, or Groq API key, you can add it in Settings and switch between local and cloud models in the same chat window. This step is entirely optional — Jan works great with local models alone.

Terminal · mac

$ Inside Jan:

$ 1. Click the Settings gear icon (bottom-left).

$ 2. Go to 'Model Providers'.

$ 3. Find the provider you want (e.g. OpenAI) and paste your API key into the field.

$ 4. Click Save.

$ 5. Back in a chat thread, open the model picker and you will now see cloud models listed alongside local ones.

What you should see

Cloud model names appear in the model picker. Selecting one routes your message to that provider's API.

This might happen

Cloud model returns an authentication error.

Double-check that you copied the full API key with no extra spaces. Keys are usually long strings starting with 'sk-'. Regenerate the key on the provider's website if needed.

// Status

cooked. baked. worked.

A fully offline AI chat app running on your own computer, with at least one local language model downloaded and ready to answer questions — no subscription, no data sent to the cloud.

// the honest bit

The honest part

Jan is a desktop app, not a web service — it only runs while the app is open on your computer. Model quality depends heavily on how much RAM you have; on 8 GB machines responses can be slow and the model may occasionally produce weaker answers than cloud ChatGPT. Very large models (13B+) require 32 GB RAM and will not run well otherwise. Jan is updated frequently, so menu names and layouts may look slightly different from this guide as new versions ship.