video benchmark
ActivityNet
A large-scale video benchmark for human activity understanding. Provides samples from 203 activity classes with an average of 137 untrimmed videos per class and 1.41 activity instances per video, for a total of 849 video hours. The benchmark covers a wide range of complex human activities that are of interest to people in their daily living and can be used to compare algorithms for three scenarios: untrimmed video classification, trimmed activity classification, and activity detection.
Updated Aug 7, 2026
Published models1
Registry coverage1
MetricScore
EvidenceB
ActivityNet leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | OP | 61.9% | 100.0% | 1 | C |
ActivityNet highlights
The top published results on this benchmark's own scale.
What is ActivityNet?
Definition and scoring fields from the benchmark registry.
A large-scale video benchmark for human activity understanding. Provides samples from 203 activity classes with an average of 137 untrimmed videos per class and 1.41 activity instances per video, for a total of 849 video hours. The benchmark covers a wide range of complex human activities that are of interest to people in their daily living and can be used to compare algorithms for three scenarios: untrimmed video classification, trimmed activity classification, and activity detection.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- ActivityNet
- Modality
- video
- Primary category
- video
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- activitynet|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about ActivityNet.
Which model scores highest on ActivityNet?
GPT-4o is currently ranked first with 61.9%.
What does ActivityNet measure?
A large-scale video benchmark for human activity understanding. Provides samples from 203 activity classes with an average of 137 untrimmed videos per class and 1.41 activity instances per video, for a total of 849 video hours. The benchmark covers a wide range of complex human activities that are of interest to people in their daily living and can be used to compare algorithms for three scenarios: untrimmed video classification, trimmed activity classification, and activity detection.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
1 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.