MMR-AD - VisA Detection: leaderboard

Zero-shot industrial anomaly detection on the VisA images of MMR-AD (2026): given a spatially aligned normal reference image and a test image, the model judges whether the test image shows an anomaly. Accuracy in percent, averaged over product classes; higher is better.

Metric: Anomaly detection accuracy (%): share of VisA test images (from MMR-AD) correctly judged normal or anomalous, zero-shot, the model sees a spatially aligned normal reference image and the test image with the MMR-AD instruction template; dataset-level average over the subdataset's product classes; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 11 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1InternVL3.5-38B74.1#1026
2O4 Mini73#372
3InternVL3-38B71.7#948
4InternVL3-8B71.5#1795
5Qwen 2.5 VL 72B70#782
6InternVL3.5-8B67.8#1618
7Qwen 2.5 VL 7B65.9#1878
8GPT-565.8#185
9Gemini 2.5 Pro65.7#280
10Llama 4 Maverick63.8#1113
11GPT-4o57.6#754

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=mmr-ad-visa-detection · How It Works · Data refreshed daily, snapshot 2026-11-22.