reasoning benchmark
EgoSchema
A diagnostic benchmark for very long-form video language understanding consisting of over 5000 human curated multiple choice questions based on 3-minute video clips from Ego4D, covering a broad range of natural human activities and behaviors
Updated Aug 7, 2026
Published models9
Registry coverage9
MetricScore
EvidenceB
EgoSchema leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | AC | 77.9% | 100.0% | 9 | C | |
| 2 | AC | 76.2% | 87.5% | 9 | C | |
| 3 | OP | 72.2% | 75.0% | 9 | C | |
| 4 | AM Nova ProAmazon | 72.1% | 62.5% | 9 | C | |
| 5 | GO | 71.5% | 50.0% | 9 | C | |
| 6 | AMNova LiteAmazon | 71.4% | 37.5% | 9 | C | |
| 7 | AC | 68.6% | 25.0% | 9 | C | |
| 8 | GO | 67.2% | 12.5% | 9 | C | |
| 9 | GO | 55.7% | 0.0% | 9 | C |
Score distribution
Top published rows on the benchmark's original scale.
EgoSchema highlights
The top published results on this benchmark's own scale.
What is EgoSchema?
Definition and scoring fields from the benchmark registry.
A diagnostic benchmark for very long-form video language understanding consisting of over 5000 human curated multiple choice questions based on 3-minute video clips from Ego4D, covering a broad range of natural human activities and behaviors
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- EgoSchema
- Modality
- video
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- egoschema|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about EgoSchema.
Which model scores highest on EgoSchema?
Qwen2-VL-72B-Instruct is currently ranked first with 77.9%.
What does EgoSchema measure?
A diagnostic benchmark for very long-form video language understanding consisting of over 5000 human curated multiple choice questions based on 3-minute video clips from Ego4D, covering a broad range of natural human activities and behaviors
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
9 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.