Instructions to use tencent/Hunyuan3D-2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Hunyuan3D-2
How to use tencent/Hunyuan3D-2 with Hunyuan3D-2:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
Benchmark Data: Hunyuan3D-2 scored 0% defect rate across 101 models in 2,307-asset census
Hi Hunyuan3D team,
I recently ran an automated, reproducible geometry benchmark (3dqa) across 2,307 AI-generated exports from 23 generators using the 3D Arena dataset on Hugging Face.
Out of all 23 generators evaluated, Hunyuan3D-2 was the only model to achieve a 0% defect rate across every single geometry check (watertightness, manifold edges, UV presence, inverted normals, floating debris) across all 101 of its evaluated assets. Hunyuan3D-2.1 followed closely at 2.0% non-watertight.
I’ve published the complete runner scripts, asset manifests, and raw aggregate reports for full transparency and reproducibility:
https://github.com/alza123123/3dqa-benchmark
If you'd like to inspect or verify the raw evaluation data for your models, you can re-run the checks via pip install 3dqa. Thanks for producing such clean mesh outputs.