MLVU
A comprehensive benchmark for multi-task long video understanding that evaluates multimodal large language models on videos ranging from 3 minutes to 2 hours across 9 distinct tasks including reasoning, captioning, recognition, and summarization.
9rows
scoreprimary metric
2026-05-06sampled
Metadata
Metrics
Score, Normalized Score
| Rank | Subject | Score | Model Match | Provenance | Sampled |
|---|---|---|---|---|---|
| 1 | Qwen3.5-122B-A10B | 0.87 |  Qwen3.5-122B-A10B qwen-qwen3.5-122b-a10b | Self-reported | 2026-05-06 |
| 2 | Qwen3.6 Plus | 0.87 | Qwen3.6 Plus qwen-qwen3.6-plus | Self-reported | 2026-05-06 |
| 3 | Qwen3.6-27B | 0.87 | Qwen3.6 27B qwen-qwen3.6-27b | Self-reported | 2026-05-06 |
| 4 | Qwen3.6-35B-A3B | 0.86 | Qwen3.6 35B A3B qwen-qwen3.6-35b-a3b | Self-reported | 2026-05-06 |
| 5 | Qwen3.5-27B | 0.86 | Qwen3.5-27B qwen-qwen3.5-27b | Self-reported | 2026-05-06 |
| 6 | Qwen3.5-35B-A3B | 0.86 | Qwen3.5-35B-A3B qwen-qwen3.5-35b-a3b | Self-reported | 2026-05-06 |
| 7 | Qwen3 VL 235B A22B Instruct | 0.84 | Qwen3 VL 235B A22B Instruct qwen-qwen3-vl-235b-a22b-instruct | Self-reported | 2026-05-06 |
| 8 | Qwen3 VL 235B A22B Thinking | 0.84 | Qwen3 VL 235B A22B Thinking qwen-qwen3-vl-235b-a22b-thinking | Self-reported | 2026-05-06 |
| 9 | Qwen2.5 VL 7B Instruct | 0.70 | — | Self-reported | 2026-05-06 |
No matching rows.