Qwen 2 VL 72B: benchmark results
Alibaba's open Qwen 2 VL 72B vision-language model. Provider: Alibaba. Released 2024-09-19. Access: Open.
Aggregate Rating 1595 ± 8, rank #460 of 1656 rated models, from 1201 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CG-Bench | 45.3 | Long-Video MCQ Accuracy (%) | 100 |
| CHOICE | 0.74 | Overall Score | 100 |
| MEGA-Bench Task - Electricity Future Prediction From Table | 77.8 | Task Score (%) | 100 |
| MEGA-Bench Task - FUNSD Document QA | 92.9 | Task Score (%) | 100 |
| MEGA-Bench Task - Funny Image Title | 75 | Task Score (%) | 100 |
| MEGA-Bench Task - MFC Bench Check Text Style | 78.6 | Task Score (%) | 100 |
| MEGA-Bench Task - Monthly Weather Days Count | 54.8 | Task Score (%) | 100 |
| MEGA-Bench Task - Multilingual Game Info Parsing | 88.4 | Task Score (%) | 100 |
| MEGA-Bench Task - Newspaper OCR In Query Box | 80 | Task Score (%) | 100 |
| MEGA-Bench Task - Panel Images Multi Question | 90.5 | Task Score (%) | 100 |
| MEGA-Bench Task - Paper VQA | 78.6 | Task Score (%) | 100 |
| Open LMM Reasoning - MMMath - Knowledge-L2-Quadratic Function | 9.6 | Accuracy (%) | 100 |
Interactive version: theaggregate.ai/model?slug=qwen-2-vl-72b · How It Works · Data refreshed daily, snapshot 2026-10-20.