Qwen 2 VL 72B Instruct: benchmark results
Alibaba's open 72B vision-language flagship (September 2024) with dynamic-resolution image handling and 20-minute-plus video understanding. Provider: Alibaba. Released 2024-09-19. Access: Open.
Aggregate Rating 1568 ± 10, rank #551 of 1635 rated models, from 169 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LLM Stats (EgoSchema) | 77.9 | Score (%) | 100 |
| Open LLM Leaderboard - MMLU-Pro | 57.17 | Score | 99.1 |
| KOFFVQA - Object Attributes | 86.67 | Score (%) | 98.8 |
| Open LLM Leaderboard - BBH | 69.46 | Score | 98.5 |
| Open LLM Leaderboard - GPQA | 38.76 | Score | 96.9 |
| LLM Stats (TextVQA) | 85.5 | Score (%) | 93.8 |
| TrustGen - HallusionBench Easy | 73.03 | Accuracy (%) | 88.9 |
| TrustGen - HallusionBench Overall | 61.69 | Accuracy (%) | 88.9 |
| CareQA-Vision - MCQ - Nursing | 64.29 | Accuracy (%; 42 image-based multiple-choice questions from E | 88.2 |
| MERA Multi - UniScienceVQA | 25.5 | Score (%) | 88.1 |
| MERA Multi - ruTiE-Image | 66.8 | Score (%) | 86.2 |
| Open LLM Leaderboard - MATH Level 5 | 34.44 | Score | 84.9 |
Interactive version: theaggregate.ai/model?slug=qwen-2-vl-72b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-05.