ZimaSpace
Store

How to Manually Download the Large Language Model

Overview

The Local LLM Inference guide pulls the model straight from Hugging Face — the fastest route when your NAS has a good connection. This page covers the offline route: download the model on a PC, then move it to the NAS over USB or your LAN.

The steps below use the same model as the main guide — Qwen3.6-35B-A3B-UD-IQ3_XXS.gguf, about 12.3 GB — and the same destination folder. The download-and-transfer pattern works for any GGUF model.

Before You Start

  • A PC with internet access
  • About 13 GB of free space on the PC, and the same on the NAS
  • A USB drive, or the NAS reachable over your LAN

Step 1: Download the Model on Your PC

The Hugging Face CLI is the easiest way — it resumes interrupted downloads:

pip install -U huggingface_hub
hf download unsloth/Qwen3.6-35B-A3B-GGUF \
  Qwen3.6-35B-A3B-UD-IQ3_XXS.gguf --local-dir models/llm

You can also download from the model page in a browser.

Step 2: Move the Model to the NAS

Over USB: copy the models/llm folder to the drive, plug it into the NAS, then in ZimaOS Files move the folder to the NAS’s models/llm directory (create it if it does not exist).

Over LAN: push the file from the PC with scp. Create the destination folder on the NAS first:

ssh <username>@<your-nas-ip> mkdir -p models/llm
scp models/llm/Qwen3.6-35B-A3B-UD-IQ3_XXS.gguf <username>@<your-nas-ip>:models/llm/

Step 3: Verify the File

Large downloads can corrupt silently. On the NAS, check the size and the checksum:

ls -lh models/llm
sha256sum models/llm/Qwen3.6-35B-A3B-UD-IQ3_XXS.gguf

Compare both against the file card on the model page — the file should be about 12.3 GB.

Step 4: Continue the Setup

The model is now in models/llm — exactly where the Local LLM Inference guide expects it. Skip that guide’s download step and continue with starting the server.

Reference Links