multimodal benchmark
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.
Updated Aug 7, 2026
Published models10
Registry coverage10
MetricScore
EvidenceB
MLVU leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | AC | 87.4% | 100.0% | 10 | C | |
| 2 | AC | 87.3% | 88.9% | 10 | C | |
| 3 | AC | 86.7% | 77.8% | 10 | C | |
| 4 | AC | 86.6% | 66.7% | 10 | C | |
| 5 | AC | 86.2% | 55.6% | 10 | C | |
| 6 | AC | 85.9% | 44.4% | 10 | C | |
| 7 | AC | 85.6% | 33.3% | 10 | C | |
| 8 | AC | 84.3% | 22.2% | 10 | C | |
| 9 | AC | 83.8% | 11.1% | 10 | C | |
| 10 | AC | 70.2% | 0.0% | 10 | C |
Score distribution
Top published rows on the benchmark's original scale.
MLVU highlights
The top published results on this benchmark's own scale.
What is MLVU?
Definition and scoring fields from the benchmark registry.
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.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- MLVU
- Modality
- multimodal
- Primary category
- multimodal
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- mlvu|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about MLVU.
Which model scores highest on MLVU?
Qwen3.7-Plus is currently ranked first with 87.4%.
What does MLVU measure?
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.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
10 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.