Qwen 3 VL 30B A3B Instruct: 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 2025-10-04. Access: Open.
Aggregate Rating 1621 ± 5, rank #384 of 1656 rated models, from 649 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| BeSafe-Bench (Embodied Planning) | 26.09 | Success-and-safe rate in percent: share of tasks completed w | 100 |
| ESPIRE - Place Execution | 43.47 | Execution acceptance rate (%) on the 1,125 place tasks: the | 100 |
| ESPIRE - Place Success | 20 | Task success rate (%) on the 1,125 place tasks: correct loca | 100 |
| EarthSpatialBench (CoT) - Within Counting | 2.83 | Mean absolute error of the object count on EarthSpatialBench | 100 |
| EarthSpatialBench (Direct) - Angle Estimation | 117.2 | Mean absolute error in degrees of the estimated azimuth betw | 100 |
| EarthSpatialBench (Direct) - Within Counting | 2.45 | Mean absolute error of the object count on EarthSpatialBench | 100 |
| FlagEval EmbodiedVerse - EgoPlan-Bench2 | 58.8 | Score | 100 |
| GAVEL - Grounding Accuracy (IoU 0.5) | 22.8 | Accuracy (%; share of predicted evidence boxes with IoU of a | 100 |
| GameplayQA - Event Recognition | 82 | Accuracy (%) on the 61 Event Recognition questions (the mode | 100 |
| GameplayQA - Occurrence Count | 65.3 | Accuracy (%) on the 75 Occurrence Count questions (the model | 100 |
| MedQ-Deg - Intensity Degradation | 76.2 | Accuracy (%) on MedQ-Deg items whose images carry intensity- | 100 |
| MemLens (32K) - Temporal Reasoning | 60.82 | Accuracy (%) on the temporal reasoning (TR, 194 questions) a | 100 |
Interactive version: theaggregate.ai/model?slug=qwen-3-vl-30b-a3b-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.