LLaVA-Video-72B-Qwen2: benchmark results
Provider: LLaVA Team. Access: Open.
Aggregate Rating 1582 ± 24, rank #1219 of 10916 rated models, from 84 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenVLM Video - MVBench - Unexpected Action | 90 | Accuracy (%) | 100 |
| OpenVLM Video - Video-MME (w/o subs) - Counting Problem | 50.4 | Accuracy (%) | 100 |
| OpenVLM Video - MLVU - Ego Reasoning | 69 | Accuracy (%) | 98.6 |
| OpenVLM Video - MLVU - Plot QA | 83.3 | Accuracy (%) | 97.2 |
| OpenVLM Video - MLVU - Needle QA | 81.4 | Accuracy (%) | 94.4 |
| OpenVLM Video - Video-MME (w/o subs) - Action Recognition | 70.6 | Accuracy (%) | 94.4 |
| OpenVLM Video - Video-MME (w/o subs) - Object Reasoning | 70.7 | Accuracy (%) | 94.4 |
| LongVideoBench | 64.9 | Test Total Accuracy (%) | 94.1 |
| OpenVLM Video - Video-MME (w/o subs) - Action Reasoning | 65.6 | Accuracy (%) | 93.1 |
| OpenVLM Video - Video-MME (w/o subs) - Long Videos | 62.1 | Accuracy (%) | 93.1 |
| OpenVLM Video - Video-MME (w/o subs) - Spatial Reasoning | 85.7 | Accuracy (%) | 93.1 |
| OpenVLM Video - MLVU | 69.98 | Normalized Score (%) | 91.7 |
Interactive version: theaggregate.ai/model?slug=llava-video-72b-qwen2 · How It Works · Data refreshed daily, snapshot 2026-11-08.