Instructions to use ccore/Qwen3.5-4B-GLB-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-GLB-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-GLB-3D-Coder:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf ccore/Qwen3.5-4B-GLB-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-GLB-3D-Coder:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf ccore/Qwen3.5-4B-GLB-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-GLB-3D-Coder:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama cli -hf ccore/Qwen3.5-4B-GLB-3D-Coder:Q4_K_M

Use Docker

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

Llama new
LM Studio
Jan

vLLM

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

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ccore/Qwen3.5-4B-GLB-3D-Coder"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ccore/Qwen3.5-4B-GLB-3D-Coder",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

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

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

How to use ccore/Qwen3.5-4B-GLB-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-GLB-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-GLB-3D-Coder:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

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

Lemonade

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

Pull the model

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

Run and chat with the model

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

List all available models

lemonade list

Hermes Agent

How to use ccore/Qwen3.5-4B-GLB-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-GLB-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-GLB-3D-Coder:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use ccore/Qwen3.5-4B-GLB-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-GLB-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-GLB-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"

ccore commited on about 3 hours ago

Commit

38bfa74

verified ·

1 Parent(s): 541152f

Streamline model card and consolidate training updates and future plans

Browse files

Files changed (1) hide show

README.md +65 -97

README.md CHANGED Viewed

@@ -18,53 +18,88 @@ tags:
 # Qwen3.5-4B-GLB-3D-Coder
-### Early preview · Create 3D scenes and assets you can export as GLB
-A small **4B coding model** being trained to generate **Three.js scenes and assets with GLB export**, so you can bring the results into editors such as **Godot**.
-This is an independent model by **CCORE**, based on [Qwen/Qwen3.5-4B](https://huggingface.co/Qwen/Qwen3.5-4B). It is not an official Qwen release.
-**Dataset release update:** The training dataset will be released tomorrow, **November 3**. Training is still in progress: I’ve extended the run to aim for a stronger checkpoint than the current preview.
-## Community showcase: A 3D RPG built with GLB-3D-Coder!
-One of the coolest things about releasing this model is seeing what people are building with it!
-**Naoki Kishida (@kis)** created a **playable 3D RPG demo using assets generated with GLB-3D-Coder**.
-It's a great example of what I had in mind when developing this model: generating reusable 3D assets that can be used in games and interactive environments.
 [![Gameplay from Naoki Kishida's 3D RPG demo, featuring assets generated with GLB-3D-Coder](images/community-rpg-naoki-kishida.gif)](https://kishida.github.io/webdemos/rpg-3d-glb/)
 [**Play the 3D RPG demo**](https://kishida.github.io/webdemos/rpg-3d-glb/)
-Huge thanks to **@kis** for experimenting with the model and sharing this project with the community!
-## Latest checkpoint update
-**Updated November 2, 2026 at 04:50 UTC.**
-Both **Q4_K_M** and **Q8_0** have been replaced with a newer checkpoint. In my latest tests, **generation quality has improved substantially** over the previous preview. The gallery below now shows new results, including the exported assets imported into **Godot**.
-**If you downloaded either GGUF before the UTC timestamp above, delete the old weights from LM Studio or whichever app you use and download them again. The filenames have not changed, so make sure you are loading the updated files.**
-**The optional vision projector is now included** for image-reference inputs; see [Optional vision support](#optional-vision-support) below.
-The **sampling settings and system prompt remain unchanged**. Training is still running, and the full Safetensors weights and training dataset will follow once it finishes.
-## Preview status and release plans
-This is an **updated preview** from the ongoing training run. Training is still in progress. Once training finishes, I plan to release the **full Safetensors weights and the training dataset**, just as I did with the previous 3D Coder release.
-In my tests, the model already generates useful scenes and assets, and **most generations work on the first attempt**, especially with **English prompts**. Some generations still need another attempt or a correction.
-I am releasing this preview so people can start experimenting while training continues. I expect quality to improve with later checkpoints, and feedback on both successful results and failures is welcome.
 ## Training domains and prompt ideas
-The training dataset for this GLB model spans **25 game-asset domains**. The examples below are **prompt ideas to try** in these areas. Use the [system prompt](#system-prompt), then send a short English request describing the asset you want.
-| Domain represented in the training data | Example prompt to try |
 |:---|:---|
 | Characters and NPCs | `Create a village blacksmith wearing a blue apron.` |
 | Enemies | `Create a swamp goblin scout with a wooden shield.` |
@@ -92,88 +127,21 @@ The training dataset for this GLB model spans **25 game-asset domains**. The exa
 | Underwater structures | `Create a coral observation dome with a small docking platform.` |
 | Puzzles and gameplay markers | `Create a glowing rune sequence console with three stone buttons.` |
-You can vary the style, setting, materials, colors, and details. For assets that need movement, explicitly request the animation; the system prompt asks for it to be included in the exported GLB.
-## From a prompt to a GLB asset
-Describe the scene or asset you want. The model generates HTML/JavaScript that builds it in Three.js, with a GLB export option. Open the generated scene, export the asset, and import the `.glb` file into your editor.
-The screenshots below show generated assets in the browser and exported assets imported into **Godot**.
-[![New checkpoint assets exported as GLB and imported into Godot](images/godot-imported-assets-01.png)](images/godot-imported-assets-01.png)
-## Example results — updated checkpoint
-**The assets below were generated by this updated GLB-3D-Coder checkpoint using prompts that were not used in its training.** These example prompts are not included in the training dataset I plan to release.
-The captions show the user prompts used for these examples. Short English prompts are enough; use the system prompt listed below.
-| Jungle siege platform | Weathered snow siege platform |
-|:---:|:---:|
-| [![Jungle siege platform generated by the updated checkpoint](images/jungle-siege-platform.png)](images/jungle-siege-platform.png) | [![Weathered snow siege platform generated by the updated checkpoint](images/snow-siege-platform.png)](images/snow-siege-platform.png) |
-| **Prompt:** `Build a jungle siege platform` | **Prompt:** `Build a weathered snow siege platform` |
-| **Blue goblin brute** | **Red orc leader** |
-| [![Blue goblin brute generated by the updated checkpoint](images/blue-goblin-brute.png)](images/blue-goblin-brute.png) | [![Red orc leader generated by the updated checkpoint](images/red-orc-leader.png)](images/red-orc-leader.png) |
-| **Prompt:** `Create a blue goblin brute` | **Prompt:** `Create a red orc Leader` |
-| **Reinforced mountain command post — result 1** | **Reinforced mountain command post — result 2** |
-| [![Reinforced mountain command post, first example](images/mountain-command-post-01.png)](images/mountain-command-post-01.png) | [![Reinforced mountain command post, second example](images/mountain-command-post-02.png)](images/mountain-command-post-02.png) |
-| **Prompt:** `Make a reinforced montain command post` | **Same prompt:** `Make a reinforced montain command post` |
-### Exported assets in Godot
-The images named **Godot** show how the exported GLB assets look after import into the Godot editor.
-[![Updated checkpoint assets after import into the Godot editor](images/godot-imported-assets-02.png)](images/godot-imported-assets-02.png)
-Click any image to view it at full size.
-## Available GGUF files
-| Quantization | File |
-|:---|:---|
-| Q8_0 | [Qwen3.5-4B-GLB-3D-Coder-Q8_0.gguf](Qwen3.5-4B-GLB-3D-Coder-Q8_0.gguf) |
-| Q4_K_M | [Qwen3.5-4B-GLB-3D-Coder-Q4_K_M.gguf](Qwen3.5-4B-GLB-3D-Coder-Q4_K_M.gguf) |
-The full Safetensors weights and training dataset will follow when training is complete.
-## Optional vision support
-For **image-reference inputs** in a runtime that supports Qwen3.5 multimodal GGUF, download the optional vision encoder/projector and load it alongside the model:
-[**Download mmproj-Qwen3.5-4B-BF16.gguf**](https://huggingface.co/ccore/Qwen3.5-4B-GLB-3D-Coder/resolve/main/mmproj-Qwen3.5-4B-BF16.gguf?download=true) — approximately **644 MiB**.
-**The same projector is used with either Q4_K_M or Q8_0.** It is an optional companion file for image inputs; it is not needed for text-only prompts.
-This is the **original, unmodified Qwen3.5-4B BF16 vision encoder/projector**, also provided with [CCORE's previous 3D Coder model](https://huggingface.co/ccore/Qwen3.5-4B-3D-Coder/blob/main/mmproj-Qwen3.5-4B-BF16.gguf). It comes from [lmstudio-community/Qwen3.5-4B-GGUF](https://huggingface.co/lmstudio-community/Qwen3.5-4B-GGUF/tree/f9f88ac3e234be915e23811a6d28ea287bdb927e), based on [Qwen/Qwen3.5-4B](https://huggingface.co/Qwen/Qwen3.5-4B), under Apache 2.0.
-## Generation settings
-These are the settings I am currently using:
-| Parameter | Value |
-|:---|:---|
-| Temperature | **0.8** |
-| Top K | **40** |
-| Top P | **0.95** |
-| Min P | **0.0** |
-This is a preview, and these settings are a starting point. Feel free to test other combinations and share your results.
-## System prompt
-Use the following as the system message, then send your asset request as the user message.
-```text
-Create game assets for the Godot editor, which uses GLB files. Create whatever the user requests as a good-quality low-poly asset using only Three.js in one complete standalone HTML file. The HTML must show the asset and include an Export GLB button that downloads a working GLB. If the requested asset needs animation, add it and include the animation in the exported GLB; otherwise make it static.
-```
-## Current limitations
-- Some generations can need a retry or a small code correction.
-- English prompts currently work better in my tests.
-- This is an early checkpoint, and behavior may change as training progresses.
-Please share what you create, which settings you use, and any issues you find. Community testing will help guide the next updates.
 ## ❤️ Support Future Open-Source AI Models

 # Qwen3.5-4B-GLB-3D-Coder
+A **4B coding model** that generates standalone **Three.js HTML/JavaScript scenes with GLB export**, for use in editors such as **Godot**.
+An independent **CCORE preview** based on [Qwen/Qwen3.5-4B](https://huggingface.co/Qwen/Qwen3.5-4B). This is not an official Qwen release.
+## Latest checkpoint update
+**Latest GGUF update: November 2, 2026 at 04:50 UTC.** Both Q4_K_M and Q8_0 were replaced with a newer checkpoint, with substantially better visual quality in my tests. Sampling settings and the system prompt are unchanged.
+**If you downloaded a GGUF before that timestamp, delete the old copy and download it again in LM Studio or your app. The filenames have not changed.**
+## Community showcase: A 3D RPG built with GLB-3D-Coder!
+**Naoki Kishida (@kis)** built a playable 3D RPG demo using assets generated with this model. It’s great to see these assets being used in a game—thanks for sharing it with the community!
 [![Gameplay from Naoki Kishida's 3D RPG demo, featuring assets generated with GLB-3D-Coder](images/community-rpg-naoki-kishida.gif)](https://kishida.github.io/webdemos/rpg-3d-glb/)
 [**Play the 3D RPG demo**](https://kishida.github.io/webdemos/rpg-3d-glb/)
+## Available GGUF files
+| Quantization | File |
+|:---|:---|
+| Q8_0 | [Qwen3.5-4B-GLB-3D-Coder-Q8_0.gguf](Qwen3.5-4B-GLB-3D-Coder-Q8_0.gguf) |
+| Q4_K_M | [Qwen3.5-4B-GLB-3D-Coder-Q4_K_M.gguf](Qwen3.5-4B-GLB-3D-Coder-Q4_K_M.gguf) |
+## Optional vision support
+For image-reference inputs in a runtime that supports Qwen3.5 multimodal GGUF, load the [**BF16 vision projector**](https://huggingface.co/ccore/Qwen3.5-4B-GLB-3D-Coder/resolve/main/mmproj-Qwen3.5-4B-BF16.gguf?download=true) (**644 MiB**) alongside either Q4_K_M or Q8_0. It is optional for text-only use.
+This is the **original, unmodified Qwen3.5-4B vision encoder/projector** from [lmstudio-community/Qwen3.5-4B-GGUF](https://huggingface.co/lmstudio-community/Qwen3.5-4B-GGUF/tree/f9f88ac3e234be915e23811a6d28ea287bdb927e), under Apache 2.0, also provided with [CCORE’s previous 3D Coder model](https://huggingface.co/ccore/Qwen3.5-4B-3D-Coder/blob/main/mmproj-Qwen3.5-4B-BF16.gguf).
+## How to use
+Use the system message below, then send a short English asset request. Save the generated HTML, open it in a browser, and click **Export GLB** to download the asset for your editor. Request an animation explicitly when needed.
+In my tests, **most generations work on the first attempt**, especially in English. Some still need a retry or a small code correction.
+### Generation settings
+| Parameter | Value |
+|:---|:---|
+| Temperature | **0.8** |
+| Top K | **40** |
+| Top P | **0.95** |
+| Min P | **0.0** |
+These are my current settings; feel free to experiment with other combinations.
+### System prompt
+```text
+Create game assets for the Godot editor, which uses GLB files. Create whatever the user requests as a good-quality low-poly asset using only Three.js in one complete standalone HTML file. The HTML must show the asset and include an Export GLB button that downloads a working GLB. If the requested asset needs animation, add it and include the animation in the exported GLB; otherwise make it static.
+```
+## Example results — updated checkpoint
+These assets were generated by this checkpoint using **prompts excluded from the model’s training data and the dataset planned for release**.
+| Jungle siege platform | Weathered snow siege platform |
+|:---:|:---:|
+| [![Jungle siege platform generated by the updated checkpoint](images/jungle-siege-platform.png)](images/jungle-siege-platform.png) | [![Weathered snow siege platform generated by the updated checkpoint](images/snow-siege-platform.png)](images/snow-siege-platform.png) |
+| **Prompt:** `Build a jungle siege platform` | **Prompt:** `Build a weathered snow siege platform` |
+| **Blue goblin brute** | **Red orc leader** |
+| [![Blue goblin brute generated by the updated checkpoint](images/blue-goblin-brute.png)](images/blue-goblin-brute.png) | [![Red orc leader generated by the updated checkpoint](images/red-orc-leader.png)](images/red-orc-leader.png) |
+| **Prompt:** `Create a blue goblin brute` | **Prompt:** `Create a red orc Leader` |
+| **Reinforced mountain command post — result 1** | **Reinforced mountain command post — result 2** |
+| [![Reinforced mountain command post, first example](images/mountain-command-post-01.png)](images/mountain-command-post-01.png) | [![Reinforced mountain command post, second example](images/mountain-command-post-02.png)](images/mountain-command-post-02.png) |
+| **Prompt:** `Make a reinforced montain command post` | **Same prompt:** `Make a reinforced montain command post` |
+### Exported assets in Godot
+The exported GLB assets after import into the Godot editor:
+| Godot import — example 1 | Godot import — example 2 |
+|:---:|:---:|
+| [![New checkpoint assets exported as GLB and imported into Godot](images/godot-imported-assets-01.png)](images/godot-imported-assets-01.png) | [![Updated checkpoint assets after import into the Godot editor](images/godot-imported-assets-02.png)](images/godot-imported-assets-02.png) |
 ## Training domains and prompt ideas
+The current preview was trained across **25 game-asset domains**. Here is one prompt idea to try for each category:
+| Domain | Prompt idea |
 |:---|:---|
 | Characters and NPCs | `Create a village blacksmith wearing a blue apron.` |
 | Enemies | `Create a swamp goblin scout with a wooden shield.` |
 | Underwater structures | `Create a coral observation dome with a small docking platform.` |
 | Puzzles and gameplay markers | `Create a glowing rune sequence console with three stone buttons.` |
+You can vary the style, setting, materials, colors, and details. Share your results, settings, and any issues you find!
+## Release status and next steps
+The model is continuing to improve, so I’ve decided to **double its training dataset**. The additional examples have just finished generating, and I’ll continue training through tomorrow before finalizing the release.
+The original dataset focused on **simple prompts**, making asset generation accessible to people without a technical background. I noticed that my [previous 3D Coder model](https://huggingface.co/ccore/Qwen3.5-4B-3D-Coder) generalized better across different prompts, so I expanded the GLB-focused data to **cover more domains and improve support for more detailed, customized requests**.
+Once training finishes, I’ll release the **complete dataset, including all the additional examples**, alongside the **full Safetensors weights**, just as I did with the previous model.
+For the next model, I’m planning a dataset of **100,000 high-quality examples**, with context lengths ranging from **8K to 128K tokens**. The target is a training run roughly **15× larger than the one used for the current model**.
+My first goal is **reliable one-shot asset generation**. After that, I plan to develop a version designed to work directly inside an **agent harness**, where it can generate, test, and refine assets as part of a broader workflow.
+Preparing this dataset and running the training will require more resources for data generation, validation, and GPU compute. **If you can contribute toward making this next training run possible, your support would be a huge help.**
 ## ❤️ Support Future Open-Source AI Models