Alibaba unveils open-source BlenderWorld benchmark for AI 3d modeling

On October 14, Alibaba's Qwen team released BlenderWorld, an open-source benchmark designed to evaluate AI's ability to create realistic 3D content within Blender. The project, licensed under MIT, is available on GitHub, Hugging Face, and ModelScope. It includes 223 tasks and 2,651 standards, covering nine categories and 22 subcategories, such as product design, architecture, and scientific visualization. The benchmark requires AI to generate editable .blend files, along with renderings, videos, or model exports, ensuring the 3D scenes are modifiable by designers. The evaluation process involves rendering from eight adaptive viewpoints, applying 2,651 binary scoring criteria, and eight overall visual quality metrics, with final scores weighted equally between compliance and visual quality. The project supports integration with any generative framework and requires Blender 4.2.10, Linux, and Python 3.10 or higher.