LLaVA-OneVision-Qwen2-72B-OV: benchmark results
Provider: LLaVA Team.
Aggregate Rating 1499 ± 38, rank #2367 of 10916 rated models, from 17 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LongVideoBench | 63.2 | Test Total Accuracy (%) | 88.2 |
| MIBench (Multimodal Interaction) - Text-Centric | 66.01 | Accuracy (%) on the text-centric samples (the answer must co | 76.2 |
| MIBench (Multimodal Interaction) - Synergy | 50.1 | Accuracy (%) on the synergy samples (both modalities are nee | 71.4 |
| MIBench-mini (Multimodal Interaction) - Text-Centric | 68.4 | Accuracy (%) on the text-centric samples (the answer must co | 64.3 |
| SAW-Bench - Route Shape | 46.15 | Accuracy (%) on the 546 route shape questions (the shape of | 63.6 |
| MIBench (Multimodal Interaction) | 57.33 | Overall accuracy (%) on the full MIBench (multiple-choice im | 57.1 |
| SAW-Bench - Spatial Affordance | 52.47 | Accuracy (%) on the 162 spatial affordance questions (whethe | 56.8 |
| MIBench-mini (Multimodal Interaction) | 56.07 | Overall accuracy (%) on MIBench-mini, the paper's subset on | 46.4 |
| MIBench-mini (Multimodal Interaction) - Synergy | 45 | Accuracy (%) on the synergy samples (both modalities are nee | 39.3 |
| MIBench (Multimodal Interaction) - Vision-Centric | 54.61 | Accuracy (%) on the vision-centric samples (the answer must | 33.3 |
| MIBench-mini (Multimodal Interaction) - Vision-Centric | 54.8 | Accuracy (%) on the vision-centric samples (the answer must | 32.1 |
| SAW-Bench - Reverse Route Plan | 24.45 | Accuracy (%) on the 229 reverse route plan questions (the mo | 25 |
Interactive version: theaggregate.ai/model?slug=llava-onevision-qwen2-72b-ov · How It Works · Data refreshed daily, snapshot 2026-11-08.