Qwen-VL-Chat: benchmark results
Provider: Alibaba. Released 2023-08-25. Access: Open.
Aggregate Rating 1331 ± 22, rank #8150 of 10916 rated models, from 673 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HELM VHELM - Blink Perception - Visual Similarity | 100 | Prefix Quasi-Exact Match (%) | 100 |
| UAVBench - Region Classification (UAVDT-S) | 82.56 | Region-level classification accuracy (%) of a boxed object, | 100 |
| OpenVLM COCO Captions | 98.9 | CIDEr | 99.5 |
| HELM VHELM - Blink Perception - Object Localization | 100 | Prefix Quasi-Exact Match (%) | 98.9 |
| OpenVLM MMT-Bench - Deepfake Detection | 70 | Score (%) | 98.3 |
| HELM VHELM - Blink Perception - Relative Depth | 94.35 | Prefix Quasi-Exact Match (%) | 97.9 |
| OpenVLM MMT-Bench - Multiple View Image Understanding | 45 | Score (%) | 94.2 |
| OpenVLM MMT-Bench - Artwork Emotion Recognition | 60 | Score (%) | 91.7 |
| HELM VHELM - Blink Reasoning - Jigsaw | 99.33 | Prefix Quasi-Exact Match (%) | 91.5 |
| UAVBench - Region Classification (AU-AIR) | 75.16 | Region-level classification accuracy (%) of a boxed object, | 90 |
| UAVBench - Region Classification (WebUAV-3M) | 18.43 | Region-level classification accuracy (%) of a boxed object, | 90 |
| OpenVLM MMT-Bench - Religious Recognition | 75 | Score (%) | 89.6 |
Interactive version: theaggregate.ai/model?slug=qwen-vl-chat · How It Works · Data refreshed daily, snapshot 2026-11-08.