multimodal benchmark
LongVideoBench
LongVideoBench is a question-answering benchmark featuring video-language interleaved inputs up to an hour long. It includes 3,763 varying-length web-collected videos with subtitles across diverse themes and 6,678 human-annotated multiple-choice questions in 17 fine-grained categories for comprehensive evaluation of long-term multimodal understanding.
Updated Aug 7, 2026
Published models4
Registry coverage4
MetricScore
EvidenceB
LongVideoBench leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | BY | 80.6% | 100.0% | 4 | C | |
| 2 | BY | 80.6% | 66.7% | 4 | C | |
| 3 | MA | 79.8% | 33.3% | 4 | C | |
| 4 | AC | 54.7% | 0.0% | 4 | C |
Score distribution
Top published rows on the benchmark's original scale.
LongVideoBench highlights
The top published results on this benchmark's own scale.
What is LongVideoBench?
Definition and scoring fields from the benchmark registry.
LongVideoBench is a question-answering benchmark featuring video-language interleaved inputs up to an hour long. It includes 3,763 varying-length web-collected videos with subtitles across diverse themes and 6,678 human-annotated multiple-choice questions in 17 fine-grained categories for comprehensive evaluation of long-term multimodal understanding.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- LongVideoBench
- Modality
- multimodal
- Primary category
- multimodal
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- longvideobench|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about LongVideoBench.
Which model scores highest on LongVideoBench?
Seed 2.1 Pro is currently ranked first with 80.6%.
What does LongVideoBench measure?
LongVideoBench is a question-answering benchmark featuring video-language interleaved inputs up to an hour long. It includes 3,763 varying-length web-collected videos with subtitles across diverse themes and 6,678 human-annotated multiple-choice questions in 17 fine-grained categories for comprehensive evaluation of long-term multimodal understanding.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
4 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.