LLaVA-OneVision-Qwen2-72B: benchmark results
Provider: LLaVA Team.
Aggregate Rating 1638 ± 45, rank #709 of 10916 rated models, from 17 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MVPBench - Video Clarity Ranking | 41.71 | MVPBench normalized proficiency score (accuracy minus chance | 100 |
| MVPBench - Daily Action Evaluation | 61.9 | MVPBench normalized proficiency score (accuracy minus chance | 96.4 |
| MVPBench - Multi-View Perception Pairing | 72.04 | MVPBench normalized proficiency score (accuracy minus chance | 96.4 |
| Vinoground (Group Score) | 21.8 | Group Score | 91.3 |
| Vinoground (Video Score) | 35.2 | Video Score | 91.3 |
| MVPBench - Olympic Sports Matching | 16 | MVPBench normalized proficiency score (accuracy minus chance | 89.3 |
| MVPBench - Video Forensic Detection | 40 | MVPBench normalized proficiency score (accuracy minus chance | 89.3 |
| Vinoground (Text Score) | 48.4 | Text Score | 87 |
| MVPBench - Caption-Free Temporal Splicing | 8.02 | MVPBench normalized proficiency score (accuracy minus chance | 85.7 |
| MVPBench - Caption-Guided Temporal Splicing | 22.41 | MVPBench normalized proficiency score (accuracy minus chance | 85.7 |
| MVPBench - Gait Recognition Matching | 25 | MVPBench normalized proficiency score (accuracy minus chance | 85.7 |
| MVPBench - Physical Common Sense | 23.47 | MVPBench normalized proficiency score (accuracy minus chance | 85.7 |
Interactive version: theaggregate.ai/model?slug=llava-onevision-qwen2-72b · How It Works · Data refreshed daily, snapshot 2026-11-08.