language benchmark
Winogrande Leaderboard
Winogrande is a large-scale dataset of 44,000 pronoun resolution problems designed to test machine commonsense reasoning.
Updated Oct 23, 2026
Models23
Model coverage23
MetricScore
EvidenceB
Winogrande Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelOP | Score87.50% | Percentile100.00% | Participants23 | EvidenceC | Evaluated |
| Rank02 | ModelNR | Score86.27% | Percentile95.45% | Participants23 | EvidenceC | Evaluated |
| Rank03 | ModelXI | Score85.60% | Percentile90.91% | Participants23 | EvidenceC | Evaluated |
| Rank04 | ModelCO | Score85.40% | Percentile86.36% | Participants23 | EvidenceC | Evaluated |
| Rank05 | ModelAC | Score85.10% | Percentile81.82% | Participants23 | EvidenceC | Evaluated |
| Rank06 | ModelNV | Score84.53% | Percentile77.27% | Participants23 | EvidenceC | Evaluated |
| Rank07 | ModelGO | Score83.70% | Percentile72.73% | Participants23 | EvidenceC | Evaluated |
| Rank08 | ModelNR | Score83.19% | Percentile68.18% | Participants23 | EvidenceC | Evaluated |
| Rank09 | ModelAC | Score82.00% | Percentile63.64% | Participants23 | EvidenceC | Evaluated |
| Rank10 | ModelMI | Score81.30% | Percentile59.09% | Participants23 | EvidenceC | Evaluated |
| Rank11 | ModelAC | Score80.80% | Percentile54.55% | Participants23 | EvidenceC | Evaluated |
| Rank12 | ModelGO | Score80.60% | Percentile50.00% | Participants23 | EvidenceC | Evaluated |
| Rank13 | ModelMA | Score76.80% | Percentile45.45% | Participants23 | EvidenceC | Evaluated |
| Rank14 | ModelMA | Score75.30% | Percentile40.91% | Participants23 | EvidenceC | Evaluated |
| Rank15 | ModelIB | Score74.40% | Percentile36.36% | Participants23 | EvidenceC | Evaluated |
| Rank16 | ModelAC | Score72.90% | Percentile31.82% | Participants23 | EvidenceC | Evaluated |
| Rank17 | ModelGO | Score71.70% | Percentile27.27% | Participants23 | EvidenceC | Evaluated |
| Rank18 | ModelGO | Score71.70% | Percentile22.73% | Participants23 | EvidenceC | Evaluated |
| Rank19 | ModelMI | Score68.50% | Percentile18.18% | Participants23 | EvidenceC | Evaluated |
| Rank20 | ModelMI | Score67.00% | Percentile13.64% | Participants23 | EvidenceC | Evaluated |
| Rank21 | ModelGO | Score66.80% | Percentile9.09% | Participants23 | EvidenceC | Evaluated |
| Rank22 | ModelGO | Score66.80% | Percentile4.55% | Participants23 | EvidenceC | Evaluated |
| Rank23 | ModelBA | Score51.30% | Percentile0.00% | Participants23 | EvidenceC | Evaluated |
Winogrande Highlights
The leading models and scores on this benchmark.
Winogrande Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for Winogrande
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
What is Winogrande?
What Winogrande measures and how its scores work.
Winogrande is an adversarial Winograd Schema Challenge dataset that uses adversarial filtering to reduce spurious biases.
It measures performance on pronoun resolution problems requiring commonsense reasoning, 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
- Winogrande
- Modality
- text
- Primary category
- language
- 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 Winogrande.
Which model scores highest on Winogrande?
GPT-4 is currently ranked first with 87.50%.
What are the top three models on Winogrande?
The current leaders are GPT-4 (87.50%), Hermes 3 405B (86.27%), and MiMo-V2.5-Pro (85.60%).
Which Winogrande 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 Winogrande results?
The fastest matched records are GPT-4 (104.00 tok/s via Azure), Command R+ (59.00 tok/s via Cohere), and Qwen2.5-Coder 32B Instruct (44.00 tok/s via DeepInfra).
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.