3D-DefectBench - Texture (Expert Labels): leaderboard

Metric: Macro MCC (Matthews correlation from -1 to 1, macro-averaged over the four texture defect types) of the model's binary per-defect verdicts against expert labels on the cells both experts agree on (129-asset expert split); the fixed c004 design: six-view oblique RGB turntable renders of each text-to-3D asset and a rubric-guided checklist prompt. Source: arxiv.org. Saturation forecast: Around September 2028. 12 models tracked.

Top models

#ModelScore
1Gemini 3.1 Flash Lite0.41
2Gemini 3.1 Pro (Preview)0.39
3Claude Opus 4.70.3
4Gemini 2.5 Pro0.28
5GPT-4o0.26
6Claude Sonnet 4.60.23
7GPT-5.40.22
8Qwen 3.5 397B A17B0.14
9GPT-5 Mini0.12
10Qwen 2.5 VL 7B0.08
11Mistral Small 3.10.07
12Claude Haiku 4.50.03

Interactive version: theaggregate.ai/benchmark?slug=3d-defectbench-texture-expert-labels · How It Works · Data refreshed daily, snapshot 2026-09-29.