LLaVA-Video-7B-Qwen2: benchmark results
Provider: LLaVA Team. Access: Open.
Aggregate Rating 1535 ± 19, rank #1895 of 10916 rated models, from 132 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenVLM Video - Video-MME (w/o subs) - Counting Problem | 49.6 | Accuracy (%) | 97.2 |
| OpenVLM Video - MVBench - Object Shuffle | 44 | Accuracy (%) | 91.7 |
| MoHallBench - Sequential Inference Hallucination Rate (Generative) | 36 | Hallucination rate (%; share of free-form descriptions of tr | 88.9 |
| OpenVLM Video - MLVU - Action Count | 45.1 | Accuracy (%) | 88.9 |
| OpenVLM Video - MLVU - Ego Reasoning | 65.6 | Accuracy (%) | 88.9 |
| OpenVLM Video - MLVU - Needle QA | 79.4 | Accuracy (%) | 88.9 |
| OpenVLM Video - Video-MME (w/o subs) - Temporal Perception | 78.2 | Accuracy (%) | 88.9 |
| OpenVLM Video - MLVU - Plot QA | 78.3 | Accuracy (%) | 86.1 |
| LongVideoBench | 62.7 | Test Total Accuracy (%) | 85.3 |
| OpenVLM Video - MVBench - Object Interaction | 82 | Accuracy (%) | 84.7 |
| OpenVLM Video - MLVU - Action Order | 60.6 | Accuracy (%) | 83.3 |
| OpenVLM Video - MVBench - Unexpected Action | 83 | Accuracy (%) | 83.3 |
Interactive version: theaggregate.ai/model?slug=llava-video-7b-qwen2 · How It Works · Data refreshed daily, snapshot 2026-11-08.