reasoning benchmark
HellaSwag Leaderboard
HellaSwag is a commonsense natural language inference dataset with questions requiring completion of sentence endings based on physical situations and everyday activities.
Updated Oct 23, 2026
Models28
Model coverage28
MetricScore
EvidenceB
HellaSwag Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelAN | Score95.40% | Percentile100.00% | Participants28 | EvidenceC | Evaluated |
| Rank02 | ModelOP | Score95.30% | Percentile96.30% | Participants28 | EvidenceC | Evaluated |
| Rank03 | ModelGO | Score93.30% | Percentile92.59% | Participants28 | EvidenceC | Evaluated |
| Rank04 | ModelNR | Score90.19% | Percentile88.89% | Participants28 | EvidenceC | Evaluated |
| Rank05 | ModelXI | Score89.80% | Percentile85.19% | Participants28 | EvidenceC | Evaluated |
| Rank06 | ModelAN | Score89.00% | Percentile81.48% | Participants28 | EvidenceC | Evaluated |
| Rank07 | ModelCO | Score88.60% | Percentile77.78% | Participants28 | EvidenceC | Evaluated |
| Rank08 | ModelNR | Score88.19% | Percentile74.07% | Participants28 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score87.60% | Percentile70.37% | Participants28 | EvidenceC | Evaluated |
| Rank10 | ModelGO | Score86.50% | Percentile66.67% | Participants28 | EvidenceC | Evaluated |
| Rank11 | ModelGO | Score86.40% | Percentile62.96% | Participants28 | EvidenceC | Evaluated |
| Rank12 | ModelAN | Score85.90% | Percentile59.26% | Participants28 | EvidenceC | Evaluated |
| Rank13 | ModelNV | Score85.58% | Percentile55.56% | Participants28 | EvidenceC | Evaluated |
| Rank14 | ModelAC | Score85.20% | Percentile51.85% | Participants28 | EvidenceC | Evaluated |
| Rank15 | ModelMI | Score83.80% | Percentile48.15% | Participants28 | EvidenceC | Evaluated |
| Rank16 | ModelMA | Score83.50% | Percentile44.44% | Participants28 | EvidenceC | Evaluated |
| Rank17 | ModelAC | Score83.00% | Percentile40.74% | Participants28 | EvidenceC | Evaluated |
| Rank18 | ModelGO | Score81.90% | Percentile37.04% | Participants28 | EvidenceC | Evaluated |
| Rank19 | ModelIB | Score80.10% | Percentile33.33% | Participants28 | EvidenceC | Evaluated |
| Rank20 | ModelGO | Score78.60% | Percentile29.63% | Participants28 | EvidenceC | Evaluated |
| Rank21 | ModelGO | Score78.60% | Percentile25.93% | Participants28 | EvidenceC | Evaluated |
| Rank22 | ModelAC | Score76.80% | Percentile22.22% | Participants28 | EvidenceC | Evaluated |
| Rank23 | ModelGO | Score72.20% | Percentile18.52% | Participants28 | EvidenceC | Evaluated |
| Rank24 | ModelGO | Score72.20% | Percentile14.81% | Participants28 | EvidenceC | Evaluated |
| Rank25 | ModelME | Score69.80% | Percentile11.11% | Participants28 | EvidenceC | Evaluated |
| Rank26 | ModelMI | Score69.40% | Percentile7.41% | Participants28 | EvidenceC | Evaluated |
| Rank27 | ModelMI | Score69.10% | Percentile3.70% | Participants28 | EvidenceC | Evaluated |
| Rank28 | ModelBA | Score33.00% | Percentile0.00% | Participants28 | EvidenceC | Evaluated |
HellaSwag Highlights
The leading models and scores on this benchmark.
HellaSwag Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for HellaSwag
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is HellaSwag?
What HellaSwag measures and how its scores work.
HellaSwag is a commonsense natural language inference dataset that uses Adversarial Filtering to create questions designed to be trivial for humans and difficult for state-of-the-art models.
It measures performance on sentence-ending completion questions based on physical situations and everyday activities, reported as Score in ratio units.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- HellaSwag
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- overall
LLM Stats. Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about HellaSwag.
Which model scores highest on HellaSwag?
Claude 3 Opus is currently ranked first with 95.40%.
What are the top three models on HellaSwag?
The current leaders are Claude 3 Opus (95.40%), GPT-4 (95.30%), and Gemini 1.5 Pro (93.30%).
Which HellaSwag model has the lowest official input price?
Phi 4 Mini has the lowest matched official input price at $0.08 input / $0.30 output per 1M tokens.
Which models are fastest among HellaSwag results?
The fastest matched records are Llama 3.2 3B Instruct (171.50 tok/s via DeepInfra), Gemini 1.5 Flash (150.00 tok/s via Google), and GPT-4 (104.00 tok/s via Azure).
Does the highest score result prove overall model quality?
No. This benchmark measures one defined capability or task. The overall LLMBoard score uses a separate aggregation across eligible benchmark evidence.