mPLUG-Owl2: benchmark results
Provider: Alibaba. Released 2023-11-07. Access: Open.
Aggregate Rating 1315 ± 33, rank #8429 of 10916 rated models, from 430 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenVLM MMT-Bench - Deepfake Detection | 70 | Score (%) | 98.3 |
| OpenVLM MMT-Bench - Depth Estimation | 85 | Score (%) | 98.1 |
| OpenVLM MMBench V1.1 CN - Identity Reasoning | 100 | Accuracy (%) | 97.9 |
| OpenVLM MMT-Bench - Geometrical Relativity | 75 | Score (%) | 96.1 |
| OpenVLM MMT-Bench - Color Constancy | 71.4 | Score (%) | 94.2 |
| OpenVLM MMT-Bench - Color Contrast | 75 | Score (%) | 91.7 |
| OpenVLM MMT-Bench - VI | 65 | Score (%) | 90.5 |
| OpenVLM MME - Celebrity | 167.9 | Score | 90.2 |
| FineBench - Object Manipulation | 78.5 | Accuracy (%; four-option multiple-choice questions on how a | 90 |
| OpenVLM COCO Captions | 35 | CIDEr | 88.1 |
| OpenVLM MMT-Bench - Human Object Interaction Recognition | 85 | Score (%) | 87.9 |
| OpenVLM MMT-Bench - Interactive Segmentation | 57.1 | Score (%) | 87.6 |
Interactive version: theaggregate.ai/model?slug=mplug-owl2 · How It Works · Data refreshed daily, snapshot 2026-11-08.