LLaVA-OneVision-72B: benchmark results
Provider: LLaVA Team. Released 2024-08-06. Access: Open.
Aggregate Rating 1547 ± 13, rank #1747 of 10916 rated models, from 872 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MEGA-Bench Task - Action Prediction | 78.6 | Task Score (%) | 100 |
| SFI-Bench - Troubleshooting | 61 | Accuracy (%) on the causal hypothesis and troubleshooting qu | 100 |
| OpenVLM MMBench V1.1 EN - Identity Reasoning | 100 | Accuracy (%) | 99.1 |
| OpenVLM CCBench - Historical Figures | 54.3 | Accuracy (%) | 98.2 |
| MEGA-Bench Task - MFC Bench Check Background Change | 92.9 | Task Score (%) | 98.1 |
| OpenVLM MMBench V1.1 CN - Social Relation | 91.2 | Accuracy (%) | 98.1 |
| OpenVLM MMBench V1.1 CN - Identity Reasoning | 100 | Accuracy (%) | 97.9 |
| MEGA-Bench Task - MMSoc HatefulMemes | 85.7 | Task Score (%) | 97.7 |
| OpenVLM MMBench V1.1 CN - Future Prediction | 86.7 | Accuracy (%) | 97.5 |
| OpenVLM MMBench V1.0 CN - CP | 88.9 | Accuracy (%) | 96.9 |
| SFI-Bench - Operation Planning | 58.4 | Accuracy (%) on the operation planning questions, four-optio | 96.9 |
| OpenVLM MMBench V1.1 CN - AR | 91 | Accuracy (%) | 96.8 |
Interactive version: theaggregate.ai/model?slug=llava-onevision-72b · How It Works · Data refreshed daily, snapshot 2026-11-08.