DefectBench (Building Facades): leaderboard
Metric: Defect identification F1 (0-1, times 100) of the primary defect types (crack, material loss, stain, external fixings) the model names per image, on DefectBench's 487 building-facade evaluation images (crack 117, material loss 102, surface stain 134, external fixings 122), zero-shot, JSON answers in one multi-turn dialogue per image; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 32B (Thinking) | 84.42 | #287 (Qwen 3 VL 32B) |
| 2 | Claude Opus 4.1 | 76.94 | #132 |
| 3 | GPT-4o | 75.65 | #333 |
| 4 | GLM-4.6V | 75.04 | #309 |
| 5 | GPT-5.2 Pro | 74.79 | #57 |
| 6 | Gemini 3 Pro (Preview) | 74.08 | #64 |
| 7 | GLM-4.5V | 73.99 | #339 |
| 8 | Seed 1.8 | 73.73 | #136 |
| 9 | Qwen 2.5 VL 72B Instruct | 72.94 | #364 |
| 10 | Qwen 3 VL 8B Instruct | 72.62 | #401 |
| 11 | Qwen 3 VL 32B Instruct | 71.66 | #276 |
| 12 | Gemini 3 Flash (Preview) | 71.56 | #78 |
| 13 | Qwen 2.5 VL 32B Instruct | 66.84 | #443 |
| 14 | Qwen 2.5 VL 7B Instruct | 47.96 | #643 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=defectbench-building-facades · How It Works · Data refreshed daily, snapshot 2026-10-11.