Qwen 3 VL 32B: benchmark results
Provider: Alibaba. Released 2025-10-21. Access: Open.
Aggregate Rating 1658 ± 4, rank #294 of 1656 rated models, from 908 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AeroGround - Region Counting | 45.3 | Accuracy (%; Spatial Understanding: count category instances | 100 |
| COMEX - Boundary Displacement Error | 0.11 | Mean boundary displacement error (0-1, native, lower is bett | 100 |
| CounterCount | 56.91 | Avg Acc (self-reported) | 100 |
| DefectBench (Building Facades) | 84.42 | Defect identification F1 (0-1, times 100) of the primary def | 100 |
| DermoBench - OOD Diagnosis - SNU134 | 69.58 | Accuracy (%; 4-way MCQA in the source dataset's own label sp | 100 |
| DocAttriBench - DocVQA | 29.3 | Overall answer correctness (%; grounded answer generation on | 100 |
| DocAttriBench - LongDocURL | 24.9 | Overall answer correctness (%; grounded answer generation on | 100 |
| DocAttriBench - MMLongBench-Doc | 29 | Overall answer correctness (%; grounded answer generation on | 100 |
| DocAttriBench - VisualWebBench | 21 | Overall answer correctness (%; grounded answer generation on | 100 |
| ERQA-Plus | 83.4 | Multiple-choice accuracy (%) over all 1,766 questions in 12 | 100 |
| ERQA-Plus - Action-Centric Reasoning | 84.5 | Multiple-choice accuracy (%) on the 323 action-centric reaso | 100 |
| ERQA-Plus - Contextual and Commonsense Reasoning | 88.6 | Multiple-choice accuracy (%) on the 297 contextual and commo | 100 |
Interactive version: theaggregate.ai/model?slug=qwen-3-vl-32b · How It Works · Data refreshed daily, snapshot 2026-10-20.