Instructions to use ccore/Qwen3.5-4B-3D-Coder with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use ccore/Qwen3.5-4B-3D-Coder with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf ccore/Qwen3.5-4B-3D-Coder:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf ccore/Qwen3.5-4B-3D-Coder:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf ccore/Qwen3.5-4B-3D-Coder:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf ccore/Qwen3.5-4B-3D-Coder:Q4_K_M

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf ccore/Qwen3.5-4B-3D-Coder:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama cli -hf ccore/Qwen3.5-4B-3D-Coder:Q4_K_M

Use Docker

docker model run hf.co/ccore/Qwen3.5-4B-3D-Coder:Q4_K_M

Llama new
LM Studio
Jan
Ollama
How to use ccore/Qwen3.5-4B-3D-Coder with Ollama:
```
ollama run hf.co/ccore/Qwen3.5-4B-3D-Coder:Q4_K_M
```
Unsloth Desktop

How to use ccore/Qwen3.5-4B-3D-Coder with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ccore/Qwen3.5-4B-3D-Coder:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "ccore/Qwen3.5-4B-3D-Coder:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use ccore/Qwen3.5-4B-3D-Coder with Docker Model Runner:
```
docker model run hf.co/ccore/Qwen3.5-4B-3D-Coder:Q4_K_M
```

Lemonade

How to use ccore/Qwen3.5-4B-3D-Coder with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull ccore/Qwen3.5-4B-3D-Coder:Q4_K_M

Run and chat with the model

lemonade run user.Qwen3.5-4B-3D-Coder-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use ccore/Qwen3.5-4B-3D-Coder with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ccore/Qwen3.5-4B-3D-Coder:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default ccore/Qwen3.5-4B-3D-Coder:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use ccore/Qwen3.5-4B-3D-Coder with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ccore/Qwen3.5-4B-3D-Coder:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "ccore/Qwen3.5-4B-3D-Coder:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Outstanding Quality

#14

by Merhabalar - opened 9 days ago

Discussion

Merhabalar

9 days ago

I don't know what you did to this small model to get it to produce results this good, but please keep going, especially with MoE models.

This obviously isn't a creeper, so technically it failed. But the fact that it generated a 3D model, animations, a UI, and had everything working on the first try is genuinely impressive.

It's running at around 70–80 tok/s on my RTX 5060 at Q8.

I've seen much larger local models struggle to even code a working UI, so this is seriously impressive. Keep it up!

ccore

Owner 9 days ago

•

edited 9 days ago

thanks a lot for testing the model and sharing your results! really appreciate the feedback!

just to give you an idea of how much these small models can improve, here's a comparison. the first image was generated by deepseek v4.1 flash, and the second by my 4b model.

i'm actually updating the checkpoints and fp16 safetensors on Qwen3.5-4B-GLB-3D-Coder today. the latest checkpoint has improved quite a lot, especially in terms of geometry and overall asset quality.

i think you guys are going to really like the new results haha

thanks again for giving it a try! it's great to see people experimenting with the model and sharing what works and what doesn't.

Merhabalar

8 days ago

i just tried Qwen3.5-4B-GLB-3D-Coder and its so funny lmao :D

prompt was: "Create a highly detailed Airbus A320 passenger airliner, depicted as a complete aircraft in a clean side three-quarter view. The aircraft should feature the distinctive narrow-body fuselage, rounded nose section, large cockpit windshield, two under-wing turbofan engines, swept wings with visible winglets, a conventional horizontal stabilizer and vertical tail fin, and accurately proportioned landing gear. Include the characteristic passenger cabin window rows along the fuselage, clearly defined emergency exits and doors, engine nacelles with realistic intake and exhaust geometry, detailed wing surfaces, flap-track fairings, control surfaces, and subtle panel lines. Use realistic aviation proportions and structural details, with a clean white aircraft body and understated commercial-airliner markings. Present the aircraft as a precise, realistic product-style depiction suitable for identifying the A320's exterior design. Use a plain presentation without extra scenery. Keep it static."

ccore

Owner 8 days ago

haha yeah, that definitely doesn't look much like an airbus a320 😂

i tried the exact same prompt here, but lowered the temperature to 0.6, and got this result. still not a perfect a320, of course, but quite a different outcome!

i think adjusting the sampling parameters a bit might help, especially when using more detailed prompts.

the two models also have slightly different focuses. glb-coder was trained mainly for lightweight, low-poly game assets and glb export, while the original 3d-coder focuses more on broader and more complex 3d scenes.

both can handle different types of requests, but their training data definitely influences how they interpret and generate things.

there's still plenty of room for improvement, but i think it's interesting how much the results can vary with just a few adjustments.

thanks for testing and sharing your feedback! it's really helpful for the next versions (:

ccore

Owner 8 days ago

"create an Airbus A320"

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment