Your first model

Goal

Download Qwen3.5-2B in the Q4_K_M version (1.28 GB) and confirm it landed in the data folder you chose, not on your system drive.

What a model is

A model is a single file that holds everything the AI learned during training. Bigger models are smarter and heavier. The number in the name is roughly how big it is: "2B" means about 2 billion settings (parameters). You do not need to remember that.

Every attendee uses the same small model, so everyone can follow every step, even on an 8 GB laptop. You can try bigger ones later (chapter 10).

What Q4_K_M means

The same model comes in several compressed versions. Q4 keeps about 4 bits per setting. Smaller and faster, slightly less clever. For laptops, Q4_K_M is the usual sweet spot.

Model	Version	Download	Licence
Qwen3.5-2B	Q4_K_M	1.28 GB	Apache-2.0

Source: unsloth/Qwen3.5-2B-GGUF on Hugging Face, based on Qwen/Qwen3.5-2B.

Download it

1. In Jan, open Hub in the left sidebar. (If the setup screen suggests a different model, choose Browse other models in the Hub.)
2. Search for the exact name unsloth/Qwen3.5-2B-GGUF. A plain Qwen3.5-2B search lists several copies and may not show this one. Pick the copy by Unsloth. Skip copies with "Uncensored" in the name.
3. Open it. Jan shows a label for your laptop:
   • Fits: go ahead.
   • May be slow: it works, slowly.
   • Won't fit: something is wrong for a 1.28 GB model. Close other apps and check chapter 02.
4. Choose the Q4_K_M version and click Download. Jan's list may round the size to 1.2 GB. Jan also downloads a small vision file (mmproj, about 0.67 GB), so expect about 2 GB in total.
5. Watch the progress in the top bar. The × cancels. If it fails, click Download again.

The file comes from huggingface.co. This is one of the moments Jan uses the internet.

Workshop tip: one download, many laptops

Venue Wi-Fi and 50 people downloading 1.28 GB each do not mix. A facilitator can download the .gguf file once, copy it to pen drives, and attendees import it: Settings → Model Providers → Llama.cpp → import the file. Put it on the data drive you chose, not the desktop.

Check it works

• The model appears in the model picker in a new chat.
• Look in your data folder: X:\Jan\llamacpp\models\ on Windows, or the folder from chapter 04 on Mac. The .gguf file is there, about 1.3 GB.
• Your system drive did not gain 1.3 GB. If it did, go back to chapter 03 step 4.

Exercise

Search the Hub for Jan-v3.5-4B and Qwen3.5-9B, but do not download them. Note the label each one gets on your laptop. That tells you how far you can upgrade later.