Skip to content

Hands-on · about an hour

Follow along. Finish with your own AI.

Pick your system and the drive for Jan’s data. Every step and path below changes to match. Do each step, check you see what it says, tick it off. No code, and nothing to sign up for. Want the reasons behind a step? Each one links to its chapter.

  1. Up nextStep 1

    Check your laptop

    Do this

    1. Open Settings → System → About. Read "Processor" and "Installed RAM". You need 8 GB or more.
    2. Press Ctrl + Shift + Esc → Performance. An NVIDIA graphics card makes it faster, but it is optional.
    3. Open File Explorer → This PC. Find a drive that is not C: with 10 GB free, and pick it above.

    You should see

    • You know your memory, your chip, and where Jan’s data will live.
    Didn’t work?
    • Old Celeron or Pentium? The processor needs AVX2 or Jan’s engine will not start. Chapter 02 shows how to check.
    Why this step? Chapter 02
  2. Step 2

    Make a home for your AI

    Do this

    1. Open File Explorer and go to your data drive.
    2. Right-click → New → Folder. Name it .
    3. That is : a folder, never the drive itself.

    You should see

    • An empty folder, off your system drive.
    Why this step? Chapter 03
  3. Step 3

    Download Jan

    Do this

    1. Open and click Download for Windows.
    2. Pick x64 for normal Intel and AMD laptops. Pick ARM64 only for Snapdragon.
    3. The file is called something like .

    You should see

    • The installer is in your Downloads folder.

    Only download from jan.ai or Jan’s GitHub releases. Never from “free download” sites.

    Why this step? Chapter 03
  4. Step 4

    Install it

    Do this

    1. Double-click the .
    2. Windows says “Windows protected your PC”? If the file came from jan.ai, click More info → Run anyway.
    3. Wait. Jan opens by itself.

    You should see

    • Jan’s window is open.
    Why this step? Chapter 03
  5. Step 5

    Let Jan set itself up

    Do this

    1. Keep the internet on. Click Let’s start.
    2. “Help improve Jan”: click Deny to keep analytics off.
    3. Jan downloads its engine. A badge says “Will use your GPU” or “Will run on CPU”. Both are fine.
    4. Wait for “Engine ready”. If it offers a model, do not download it yet.

    You should see

    • Engine ready.
    Didn’t work?
    • Stuck on the engine download? Check the internet, then click Re-run checks.
    Why this step? Chapter 03
  6. Step 6

    Move Jan’s data off the system drive

    Do this

    1. In Jan: Settings → General. Find App Data.
    2. Click Change Location and pick .
    3. Confirm. Jan copies everything there and restarts.
    4. Setup screen again? That is normal. Go through it once more.

    You should see

    • App Data shows .
    • now has folders like , and .

    Do this before any model download, while the data is still small. Jan copies, so an old copy stays behind; chapter 03 shows how to clean it up safely.

    Didn’t work?
    • “Cannot relocate data folder to root directory”: you picked the drive itself. Pick the folder inside it.
    • Still failing? Quit Jan, open in Notepad, set , save and reopen.
    Why this step? Chapter 03
  7. Step 7

    Download your first model

    Do this

    1. Open Hub in the left sidebar. Setup suggested another model? Choose Browse other models in the Hub.
    2. Search the exact name (a plain search may not show it). Skip any copy named “Uncensored”. Your laptop’s label should say Fits.
    3. Choose Q4_K_M (1.28 GB, Jan may show 1.2 GB) and click Download. Jan also fetches a small vision file, so expect about 2 GB in total. This part uses the internet: the files come from huggingface.co.

    You should see

    • Qwen3.5-2B appears in the model picker of a new chat.
    • The file, about 1.3 GB, is in .
    • Your C: drive did not grow by 1.3 GB.
    Didn’t work?
    • Won’t fit for a 1.28 GB model? Close other apps and check your memory again.
    • Download failed? Click Download again.
    Why this step? Chapter 05
  8. Step 8

    First chat, with Reasoning off

    Do this

    1. Click New Chat and pick Qwen3.5-2B.
    2. Set Reasoning to Off. Qwen3.5 thinks out loud by default, which is slow for chat.
    3. Ask:
    4. While it answers, open Task Manager → Performance. CPU or GPU jumps, and memory goes up: the model now sits in your RAM.

    You should see

    • The answer appears word by word. The first one is slower while the model loads.
    Didn’t work?
    • Very slow? Plug in the charger, close other apps, and check Reasoning is Off.
    Why this step? Chapter 06
  9. Step 9

    Pull the plug

    Do this

    1. Taskbar network icon → turn on Airplane mode. Unplug any cable.
    2. Load any website in your browser. It should fail.
    3. In Jan, start a New Chat and ask:
    4. Quit Jan fully (also from the tray, the ^ near the clock), open it again, and chat once more.

    You should see

    • It answers with the internet off, and your chats survived a restart.

    This proves the AI runs on your laptop. It does not prove Jan never goes online: setup, downloads and the update check do.

    Why this step? Chapter 07
  10. Step 10

    Find your chats on disk

    Do this

    1. Open in File Explorer.
    2. Open any chat folder, then open in Notepad.

    You should see

    • Your conversation, in plain text, on your own disk.

    Plain text means not encrypted. Lock your laptop, and turn on Device encryption or BitLocker, for the data drive too.

    Why this step? Chapter 08
  11. Step 11

    Check the privacy switches

    Do this

    1. Settings → Privacy → Analytics: off, unless you chose to help.
    2. Settings → General → Automatic Update Check: we suggest leaving it on. Updates fix security problems.
    3. To keep chats local, never add a cloud API key under Settings → Model Providers.
    4. Turn the internet back on.

    You should see

    • You know when Jan goes online: setup, model downloads, the update check, and analytics only if you said yes.
    Why this step? Chapter 08