Qwen 2.5 VL 72B: benchmark results
Provider: Alibaba. Released 2025-01-26. Access: Open.
Aggregate Rating 1626 ± 8, rank #790 of 10916 rated models, from 818 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MM-IFEval | 73.68 | Overall Accuracy (%) | 100 |
| MMR-AD - MPDD Localization | 24.7 | Anomaly localization accuracy (%) on MPDD test images (from | 100 |
| OVIBench (Multiple Choice) - False Trigger | 95.37 | Answer-selection accuracy (%; choosing, among four candidate | 100 |
| OmniHandwritingOCR - Hard Multi-Line Formulas | 55.81 | 1-NED (%; one minus the normalized Levenshtein distance betw | 100 |
| Open LMM Reasoning - WeMath - Direction | 100 | Accuracy (%) | 100 |
| OpenVLM A-Bench | 81 | Accuracy (%) | 100 |
| OpenVLM MMBench V1.0 CN - FP-S | 96.2 | Accuracy (%) | 100 |
| OpenVLM MMBench V1.0 EN - FP-S | 95.2 | Accuracy (%) | 100 |
| OpenVLM MMBench V1.1 CN - FP-S | 96.3 | Accuracy (%) | 100 |
| OpenVLM MMBench V1.1 EN - Image Scene | 93.8 | Accuracy (%) | 100 |
| OpenVLM Q-Bench - Yes-or-No (In-context Distortion) | 88.9 | Accuracy (%) | 100 |
| OpenVLM SEED-Bench 2 Plus | 73.8 | Accuracy (%) | 100 |
Interactive version: theaggregate.ai/model?slug=qwen-2-5-vl-72b · How It Works · Data refreshed daily, snapshot 2026-11-08.