Qwen 3 VL 235B A22B: benchmark results
Can locate interface elements for GUI agents, read text in images and generate code from visual layouts. Video training aligns events with timestamps for locating moments in a clip. Provider: Alibaba. Released 2025-09-23. Access: Open.
Aggregate Rating 1684 ± 4, rank #238 of 1656 rated models, from 930 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ADeptS-Bench - Task Success (Mobile) | 83 | Task success rate (%; share of the 358 benign mobile tasks c | 100 |
| AirNav | 5.21 | Success rate (%; test-unseen split, share of episodes that s | 100 |
| CArtBench - CuratorQA | 84 | Exact-match accuracy (%) on all 14,421 CuratorQA questions o | 100 |
| CArtBench - CuratorQA (Calligraphy) | 85 | Exact-match accuracy (%) on the calligraphy CuratorQA questi | 100 |
| CArtBench - CuratorQA (Ceramics) | 79 | Exact-match accuracy (%) on the ceramics CuratorQA questions | 100 |
| CArtBench - CuratorQA (Composition and Format) | 88 | Exact-match accuracy (%) on the composition and format (QA3, | 100 |
| CArtBench - CuratorQA (Iconography) | 81 | Exact-match accuracy (%) on the visible iconography and elem | 100 |
| CArtBench - CuratorQA (Painting) | 87 | Exact-match accuracy (%) on the painting CuratorQA questions | 100 |
| CArtBench - CuratorQA (Scene Classification) | 94 | Exact-match accuracy (%) on the scene and activity classific | 100 |
| CArtBench - CuratorQA (Sculpture) | 84 | Exact-match accuracy (%) on the sculpture CuratorQA question | 100 |
| CArtBench - CuratorQA (Stationery) | 78 | Exact-match accuracy (%) on the stationery CuratorQA questio | 100 |
| CArtBench - CuratorQA (Style to Period Inference) | 56 | Exact-match accuracy (%) on the style-to-period inference (Q | 100 |
Interactive version: theaggregate.ai/model?slug=qwen-3-vl-235b-a22b · How It Works · Data refreshed daily, snapshot 2026-10-20.