Qwen 3 VL Plus: benchmark results
Provider: Alibaba. Access: API.
Aggregate Rating 1670 ± 10, rank #504 of 10916 rated models, from 299 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| DiffCap-Bench - Action and Pose | 75.8 | Macro-averaged F1* (%) on changes in actions and body poses: | 100 |
| DiffCap-Bench - Attribute | 82.4 | Macro-averaged F1* (%) on attribute modifications such as si | 100 |
| DiffCap-Bench - Expression | 79.1 | Macro-averaged F1* (%) on facial expression changes: harmoni | 100 |
| DiffCap-Bench - Text and Symbol | 90.5 | Macro-averaged F1* (%) on added, removed or modified text an | 100 |
| EgoCoT-Bench - Reasoning Score | 3.08 | Reasoning score (out of 5): a Qwen-Max judge rates each gene | 100 |
| FlagEval EmbodiedVerse - Where2Place | 59.12 | Score | 100 |
| GUI-RobustEval - Error Depth 0 | 55.1 | Post-error success rate (%; the agent takes over a desktop t | 100 |
| GUI-RobustEval - Error Depth 1 | 46.1 | Post-error success rate (%; the agent takes over a desktop t | 100 |
| GUI-RobustEval - Error Depth 3 | 36.9 | Post-error success rate (%; the agent takes over a desktop t | 100 |
| GUI-RobustEval - Error Depth 5 | 33.5 | Post-error success rate (%; the agent takes over a desktop t | 100 |
| GraphVerse - Graph Edit Distance (Paired Images) - Accuracy | 44 | Answer accuracy (%; testmini) | 100 |
| GraphVerse - Graph Edit Distance (Paired Images) - VGR-Score | 43.3 | Visual graph reasoning score (0-100; 0.7 process + 0.3 answe | 100 |
Interactive version: theaggregate.ai/model?slug=qwen-3-vl-plus · How It Works · Data refreshed daily, snapshot 2026-11-08.