Qwen2.5-VL Instruct (7B): benchmark results
Provider: Alibaba. Access: Open.
Aggregate Rating 1600 ± 24, rank #1011 of 10916 rated models, from 175 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HELM VHELM - Exams V - Chinese (Natural Science) | 78.25 | Prefix Quasi-Exact Match (%) | 97.9 |
| HELM VHELM - Mmmu - Music | 46.67 | Prefix Quasi-Exact Match (%) | 95.7 |
| HELM VHELM - Crossmodal 3600 - English Locale | 3.58 | Prometheus Vision Rating (1-5) | 93.6 |
| HELM VHELM - Exams V - German (Natural Science) | 79.31 | Prefix Quasi-Exact Match (%) | 93.6 |
| HELM VHELM - Flickr30K | 3.03 | Prometheus Vision rating | 91.5 |
| HELM Robo-Reward-Bench - Nyu Rot Dataset Converted Externally To Rlds | 1.11 | Absolute error | 89.5 |
| HELM VHELM - Crossmodal 3600 - Maori Locale | 3.77 | Prometheus Vision Rating (1-5) | 89.4 |
| HELM VHELM - Exams V - Chinese (Social Sciences) | 76.32 | Prefix Quasi-Exact Match (%) | 89.4 |
| HELM VHELM - Bingo - T2I | 3.53 | Prometheus Vision Rating (1-5) | 88.3 |
| HELM VHELM - Mm Star Reasoning - Logical Reasoning | 78.8 | Prefix Quasi-Exact Match (%) | 88.3 |
| HELM VHELM - Viz Wiz | 78.3 | PEM | 87.2 |
| HELM VHELM - Blink Perception - Art Style | 95.73 | Prefix Quasi-Exact Match (%) | 86.2 |
Interactive version: theaggregate.ai/model?slug=qwen2-5-vl-instruct-7b · How It Works · Data refreshed daily, snapshot 2026-11-08.