reasoning benchmark
HellaSwag
A challenging commonsense natural language inference dataset that uses Adversarial Filtering to create questions trivial for humans (>95% accuracy) but difficult for state-of-the-art models, requiring completion of sentence endings based on physical situations and everyday activities
Updated Aug 7, 2026
Published models27
Registry coverage27
MetricScore
EvidenceB
HellaSwag 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.
HellaSwag highlights
The top published results on this benchmark's own scale.
What is HellaSwag?
Definition and scoring fields from the benchmark registry.
A challenging commonsense natural language inference dataset that uses Adversarial Filtering to create questions trivial for humans (>95% accuracy) but difficult for state-of-the-art models, requiring completion of sentence endings based on physical situations and everyday activities
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- HellaSwag
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- hellaswag|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about HellaSwag.
Which model scores highest on HellaSwag?
Claude 3 Opus is currently ranked first with 95.4%.
What does HellaSwag measure?
A challenging commonsense natural language inference dataset that uses Adversarial Filtering to create questions trivial for humans (>95% accuracy) but difficult for state-of-the-art models, requiring completion of sentence endings based on physical situations and everyday activities
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
27 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is marked as eligible for the current LLMBoard capability methodology.