Dataset · ccore
3D Asset Generation in GLB Format via Three.js Code
ccore/ThreeJS-3D-GLB-Assets
About 4,000 verified examples of generating Three.js 3D models with reasoning and renders.
What's inside
The dataset contains 3,918 examples from 40 subject domains in English—from characters and vehicles to machinery and furniture. Each example includes a complete dialog, thinking traces, HTML/JavaScript code for Three.js, and a result render. All objects were tested, rendered, and validated before exporting to the GLB format.
What you can do with it
- SFT training of 3D scene generation models and models via Three.js code
- Training multimodal models to link code, 3D geometry, and visual render
- Tasks
- Text generation
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 5.4K
- Likes
- 83
Models trained on it
- ccore/Qwen3.5-4B-GLB-3D-CoderText generation 69.7K
- ccore/Qwen3-1.7B-GLB-3D-CoderText generation 6.8K
- prithivMLmods/Qwen3.5-4B-GLB-3D-Coder-MTP-RepackedImage text → text 4.5K
- mradermacher/Qwen3.5-4B-GLB-3D-Coder-GGUF 3.8K
- prithivMLmods/Qwen3.5-4B-GLB-3D-Coder-GGUFImage text → text 3.2K
- mradermacher/Qwen3.5-4B-GLB-3D-Coder-MTP-Repacked-GGUF 2.4K
- mradermacher/Qwen3.5-4B-GLB-3D-Coder-MTP-Repacked-i1-GGUF 2K
- mradermacher/GEMMA-3D-CODER-GGUF 1.6K
How to load it
from datasets import load_dataset
ds = load_dataset("ccore/ThreeJS-3D-GLB-Assets")