multimodal benchmark
LVBench
LVBench is an extreme long video understanding benchmark designed to evaluate multimodal models on videos up to two hours in duration. It contains 6 major categories and 21 subcategories, with videos averaging five times longer than existing datasets. The benchmark addresses applications requiring comprehension of extremely long videos.
Updated Aug 7, 2026
Published models24
Registry coverage24
MetricScore
EvidenceB
LVBench leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
Score distribution
Top published rows on the benchmark's original scale.
LVBench highlights
The top published results on this benchmark's own scale.
What is LVBench?
Definition and scoring fields from the benchmark registry.
LVBench is an extreme long video understanding benchmark designed to evaluate multimodal models on videos up to two hours in duration. It contains 6 major categories and 21 subcategories, with videos averaging five times longer than existing datasets. The benchmark addresses applications requiring comprehension of extremely long videos.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- LVBench
- Modality
- multimodal
- Primary category
- multimodal
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- lvbench|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about LVBench.
Which model scores highest on LVBench?
Qwen3.8 Max is currently ranked first with 81.8%.
What does LVBench measure?
LVBench is an extreme long video understanding benchmark designed to evaluate multimodal models on videos up to two hours in duration. It contains 6 major categories and 21 subcategories, with videos averaging five times longer than existing datasets. The benchmark addresses applications requiring comprehension of extremely long videos.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
24 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.