mPLUG-Owl3: benchmark results
Provider: Alibaba.
Aggregate Rating 1486 ± 32, rank #2587 of 10916 rated models, from 85 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Video-MME - Long Videos (w/ subs) | 64.5 | Accuracy (%) | 70 |
| OpenVLM Video - MVBench - Action Count | 49 | Accuracy (%) | 63.9 |
| SceneBench - Inverse VQA | 40.7 | Accuracy (%) on SceneBench's 485 inverse vqa items (choosing | 61.1 |
| OpenVLM Video - MVBench - Fine-grained Action | 48 | Accuracy (%) | 59.7 |
| OpenVLM Video - MLVU - Action Count | 30.6 | Accuracy (%) | 51.4 |
| OpenVLM Video - MVBench - Object Shuffle | 39 | Accuracy (%) | 50 |
| SceneBench - Clip QA | 54.1 | Accuracy (%) on SceneBench's 368 clip qa items (questions ab | 50 |
| SceneBench - Comment Prediction | 74.1 | Accuracy (%) on SceneBench's 297 comment prediction items (c | 50 |
| Video-MME - Medium Videos (w/ subs) | 66.9 | Accuracy (%) | 49 |
| OpenVLM Video - MMBench-Video - Attribute Reasoning | 1.55 | GPT-4 Rating (0-3) | 47.7 |
| OpenVLM Video - MMBench-Video - Cross-Instance Fine-grained Perception | 1.18 | GPT-4 Rating (0-3) | 46.6 |
| OpenVLM Video - MMBench-Video - Relation Reasoning | 1.42 | GPT-4 Rating (0-3) | 45.5 |
Interactive version: theaggregate.ai/model?slug=mplug-owl3 · How It Works · Data refreshed daily, snapshot 2026-11-08.