Qwen3.5-4B-GLB-3D-Coder-GGUF

Qwen3.5-4B-GLB-3D-Coder-GGUF

Static GGUF quantizations of ccore/Qwen3.5-4B-GLB-3D-Coder, a 4.2B model, in sizes from 2.0 GB to 8.5 GB, plus two small multi-modal supplement files. Made by mradermacher.

Made by mradermacher on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: This repository holds static GGUF quantizations of ccore/Qwen3.5-4B-GLB-3D-Coder, a 4.2B model. Quantizing shrinks the weights so the model fits on smaller hardware, at some cost in quality. The list runs from 2.0 GB up to 8.5 GB, and two small files supply the multi-modal side. The name suggests a model that writes code for 3D GLB files, but the card does not say, so check the base model's page.

Who it suits: It suits people who run models locally and want the base model in a size that fits their memory, and tinkerers who compare quantization levels.

Industry | Who uses it | How they use it | Example
Local AI | Hobbyists with modest hardware | Run a 4.2B model locally in a smaller form | Pick the Q4_K_M file that the card recommends
Software development | Developers who test coding models | Try a compact model in a GGUF runner | Load a file and send it a short coding prompt
Model research | People who compare quantization | Weigh quality against size across types | Compare Q5_K_M with Q6_K on the same prompt

If you want the original full model, or weighted quants, use their own repositories. The card says weighted quants were not planned at the time.

Where to use it:

• Local runners that load GGUF files
• Computers with room for the file you choose, from about 2.0 GB to 8.5 GB
• Windows, macOS or Linux with Python, to download the files

Use case | Who | What they get back
Local testing | Hobbyists | A GGUF file you load in your runner
Size comparison | Researchers | Many quant types of one model to compare
Multi-modal use | Developers | A supplement file that goes with the model

How to set it up: What you need first:

• Python and the Hugging Face command line
• A runner that can load GGUF files
• Free space for the file you pick, between 2.0 and 8.5 GB

1. Choose a quant type from the table below.
2. Download that one file from the repository's file list.
3. Load it in your GGUF runner.

pip install -U huggingface_hub
hf download mradermacher/Qwen3.5-4B-GLB-3D-Coder-GGUF

That command fetches every file, so for a single quant pick the file from the file list on the model page instead.

How to use it: The card gives no run command. It points to TheBloke's READMEs for how to use GGUF files, including how to join files that come in several parts. The list is sorted by size, not by quality.

Title: Selected quants
Type | Size in GB | Card's note
Q2_K | 2.0 | None given
Q3_K_M | 2.4 | Lower quality
Q4_K_S | 2.7 | Fast, and recommended
Q4_K_M | 2.8 | Fast, and recommended
Q6_K | 3.6 | Good quality, says the card
Q8_0 | 4.6 | Fast, with the best quality
f16 | 8.5 | Called overkill by the card

Part | What it does
mmproj files | Multi-modal supplements, 0.5 GB and 0.8 GB
Static quants | Plain quantized versions of the base model
Model requests | The author takes requests for other quants through a Community Discussion

Example: You pick the Q4_K_M file at 2.8 GB, which the card recommends, and load it in your GGUF runner. If you need the multi-modal side, you also download one of the mmproj files.

For newcomers and experts: A good first try is Q4_K_M, since the card recommends it and it is small. If you have room to spare, try Q6_K or Q8_0 and compare the answers.

For experts, these are static quants of ccore/Qwen3.5-4B-GLB-3D-Coder. The card says the author may add weighted quants later, and that you can ask for them in a Community Discussion. It adds that IQ types are often a better pick than a non-IQ type of similar size, and it points to a graph that compares lower-quality quant types. The f16 file is the largest at 8.5 GB and the card calls it overkill.

ccore/Qwen3.5-4B-GLB-3D-Coder is the base model
It is converted and quantized into static files
You pick a size from the list
Your runner loads the GGUF file

Good to know:

Title: File size in GB
Q2_K: 2.0 GB
Q3_K_M: 2.4 GB
Q4_K_S: 2.7 GB
Q4_K_M: 2.8 GB
Q5_K_M: 3.2 GB
Q6_K: 3.6 GB
Q8_0: 4.6 GB
f16: 8.5 GB

• Order: the list is sorted by size, not by quality.
• Quality: the card marks Q3_K_M as lower quality and Q8_0 as the best.
• Weighted quants: not available at the time, and you can request them.
• Base model: ccore/Qwen3.5-4B-GLB-3D-Coder, whose own page you should read for its intended use.
• License: apache-2.0
• Cost to run: no API key or paid service named. You need memory for the file you pick.
• Data it reaches: it downloads files from Hugging Face.
• Not for: expecting full original quality from the smallest files.

On Hugging Face, 13 November 2026

• Library: transformers
• Parameters: 4.2B
• Weights: about 2.1 GB at 4-bit (estimate)
• Formats: GGUF
• Downloads in the last 30 days: 2,283
• Likes: 5
• License: apache-2.0
• Languages: en
• Based on: ccore/Qwen3.5-4B-GLB-3D-Coder
• Tags: transformers, gguf, qwen3_5, code, threejs, 3d, godot, conversational

Links

• Model: https://huggingface.co/mradermacher/Qwen3.5-4B-GLB-3D-Coder-GGUF
• Files: https://huggingface.co/mradermacher/Qwen3.5-4B-GLB-3D-Coder-GGUF/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download mradermacher/Qwen3.5-4B-GLB-3D-Coder-GGUF

Then follow the usage in its model card. Nothing in this kit runs by itself.