Qwen3-VL-Flash: benchmark results
Provider: Alibaba.
Aggregate Rating 1637 ± 24, rank #713 of 10916 rated models, from 74 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Appear2Meaning - Culture | 36.7 | Culture attribute accuracy (%, times 100; judged correct) on | 100 |
| Appear2Meaning - Origin | 24.1 | Origin attribute accuracy (%, times 100; judged correct) on | 100 |
| Appear2Meaning - Period | 32.8 | Period attribute accuracy (%, times 100; judged correct) on | 100 |
| GraphVerse - Graph Coloring - Accuracy | 50 | Answer accuracy (%; testmini) | 100 |
| ProAct-75 - Trigger Detection | 70.35 | Trigger detection macro-F1 (%, 0-100): whether the current w | 100 |
| V2X-QA - RSU Global Layout | 93.8 | Accuracy (%) on the infrastructure-side RSU Global Layout ta | 100 |
| GUI-RobustEval - Error Depth 0 | 54.9 | Post-error success rate (%; the agent takes over a desktop t | 90 |
| GUI-RobustEval - Error Depth 1 | 44.6 | Post-error success rate (%; the agent takes over a desktop t | 90 |
| V2X-QA - Ego Prediction | 55.8 | Accuracy (%) on the vehicle-side Ego Prediction task (VS4); | 88.9 |
| Appear2Meaning - Title | 53.9 | Title attribute accuracy (%, times 100; judged correct) on t | 87.5 |
| GraphVerse - Graph Coloring - VGR-Score | 60.6 | Visual graph reasoning score (0-100; 0.7 process + 0.3 answe | 83.3 |
| GraphVerse - Maximum Common Subgraph (Paired Images) - Accuracy | 3 | Answer accuracy (%; testmini) | 83.3 |
Interactive version: theaggregate.ai/model?slug=qwen3-vl-flash · How It Works · Data refreshed daily, snapshot 2026-11-08.