llmboard.aiAI model intelligence
Home

Model Rankings

OverallOpen ModelsAgentCodingReasoningMathKnowledgeInstruction FollowingTextVision
Image GenerationImage Editing
Video GenerationImage to VideoVideo Editing
Text to SpeechSpeech to Text
Embeddings

Efficiency

Chat Token PricingImage PricingVideo PricingAudio Pricing
Chat Speed & LatencyProvider Reliability

Benchmarks

GPQAMMLU-ProAIME 2025SWE-Bench VerifiedMMLUHumanity's Last ExamLiveCodeBenchMATHHumanEvalMMMU-Pro
All Benchmarks

Tools

Model DirectoryCompare Models

Scoring & Data

Scoring & Data
393 models668 benchmarks

Leaderboard Center

Overall RankingCodingCore BenchmarksPrice & ValueRuntime Performance

Modalities

All ModelsImage GenerationImage EditingVideo GenerationImage-to-VideoVideo EditingText-to-SpeechSpeech-to-TextEmbeddings

Data & Methods

Scoring MethodAll BenchmarksReasoningMath

Vendors

All VendorsOpenAIAnthropicGoogle
llmboard.aiCopyright 2026 llmboard.ai

reasoning benchmark

Winogrande Leaderboard

WinoGrande: An Adversarial Winograd Schema Challenge at Scale. A large-scale dataset of 44,000 pronoun resolution problems designed to test machine commonsense reasoning. Uses adversarial filtering to reduce spurious biases and provides a more robust evaluation of whether AI systems truly understand commonsense or exploit statistical shortcuts. Current best AI methods achieve 59.4-79.1% accuracy, significantly below human performance of 94.0%.

Updated Aug 17, 2026

Models22
Model coverage22
MetricScore
EvidenceB

On this page

  • Ranking
  • Highlights
  • Distribution
  • Top models
  • About
  • FAQ

Winogrande Ranking

Higher score ranks better on this benchmark.

22 rows
Columns

Show columns

Sort by
Rank
Model
Score
Percentile
Participants
Evidence
Evaluated
Rank01ModelOPGPT-4OpenAIScore87.5%Percentile100.0%Participants22EvidenceCEvaluatedAug 17, 2026
Rank02ModelXIMiMo-V2.5-ProXiaomiScore85.6%Percentile95.2%Participants22EvidenceCEvaluatedAug 17, 2026
Rank03ModelCOCommand R+CohereScore85.4%Percentile90.5%Participants22EvidenceCEvaluatedAug 17, 2026
Rank04ModelACQwen2 72B InstructAlibaba Cloud / Qwen TeamScore85.1%Percentile85.7%Participants22EvidenceCEvaluatedAug 17, 2026
Rank05ModelNVLlama 3.1 Nemotron 70B InstructNVIDIAScore84.5%Percentile81.0%Participants22EvidenceCEvaluatedAug 17, 2026
Rank06ModelGOGemma 2 27BGoogleScore83.7%Percentile76.2%Participants22EvidenceCEvaluatedAug 17, 2026
Rank07ModelNRHermes 3 70BNous ResearchScore83.2%Percentile71.4%Participants22EvidenceCEvaluatedAug 17, 2026
Rank08ModelACQwen2.5 32B InstructAlibaba Cloud / Qwen TeamScore82.0%Percentile66.7%Participants22EvidenceCEvaluatedAug 17, 2026
Rank09ModelMIPhi-3.5-MoE-instructMicrosoftScore81.3%Percentile61.9%Participants22EvidenceCEvaluatedAug 17, 2026
Rank10ModelACQwen2.5-Coder 32B InstructAlibaba Cloud / Qwen TeamScore80.8%Percentile57.1%Participants22EvidenceCEvaluatedAug 17, 2026
Rank11ModelGOGemma 2 9BGoogleScore80.6%Percentile52.4%Participants22EvidenceCEvaluatedAug 17, 2026
Rank12ModelMAMistral NeMo InstructMistral AIScore76.8%Percentile47.6%Participants22EvidenceCEvaluatedAug 17, 2026
Rank13ModelMAMinistral 8B InstructMistral AIScore75.3%Percentile42.9%Participants22EvidenceCEvaluatedAug 17, 2026
Rank14ModelIBGranite 3.3 8B BaseIBMScore74.4%Percentile38.1%Participants22EvidenceCEvaluatedAug 17, 2026
Rank15ModelACQwen2.5-Coder 7B InstructAlibaba Cloud / Qwen TeamScore72.9%Percentile33.3%Participants22EvidenceCEvaluatedAug 17, 2026
Rank16ModelGOGemma 3n E4BGoogleScore71.7%Percentile28.6%Participants22EvidenceCEvaluatedAug 17, 2026
Rank17ModelGOGemma 3n E4B Instructed LiteRT PreviewGoogleScore71.7%Percentile23.8%Participants22EvidenceCEvaluatedAug 17, 2026
Rank18ModelMIPhi-3.5-mini-instructMicrosoftScore68.5%Percentile19.1%Participants22EvidenceCEvaluatedAug 17, 2026
Rank19ModelMIPhi 4 MiniMicrosoftScore67.0%Percentile14.3%Participants22EvidenceCEvaluatedAug 17, 2026
Rank20ModelGOGemma 3n E2BGoogleScore66.8%Percentile9.5%Participants22EvidenceCEvaluatedAug 17, 2026
Rank21ModelGOGemma 3n E2B Instructed LiteRT (Preview)GoogleScore66.8%Percentile4.8%Participants22EvidenceCEvaluatedAug 17, 2026
Rank22ModelBAERNIE 4.5BaiduScore51.3%Percentile0.0%Participants22EvidenceCEvaluatedAug 17, 2026

Winogrande Highlights

The leading models and scores on this benchmark.

Rank #1GPT-487.5%Rank #2MiMo-V2.5-Pro85.6%Rank #3Command R+85.4%Rank #4Qwen2 72B Instruct85.1%

Winogrande Score Distribution

A closer view of the leading scores on this benchmark.

Winogrande

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.

Ranking basisThis winogrande AI model leaderboard uses descending score in the benchmark's original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.

  1. 01
    OP
    GPT-4OpenAI
    Score
    87.5%
    Price
    $30 input / $60 output per 1M tokens
    Speed
    Up to 104 tok/s via Azure

    Strengths

    • Ranks #1 of 22 compared models
    • 100th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures Winogrande, not total model capability
  2. 02
    XI
    MiMo-V2.5-ProXiaomi
    Score
    85.6%
    Price
    $0.43 input / $0.87 output per 1M tokens
    Speed
    Up to 69 tok/s via Novita

    Strengths

    • Ranks #2 of 22 compared models
    • 95th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures Winogrande, not total model capability
  3. 03
    CO
    Command R+Cohere
    Score
    85.4%
    Price
    $2.5 input / $10 output per 1M tokens
    Speed
    Up to 100 tok/s via Bedrock

    Strengths

    • Ranks #3 of 22 compared models
    • 90th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures Winogrande, not total model capability
  4. 04
    AC
    Qwen2 72B InstructAlibaba Cloud / Qwen Team
    Score
    85.1%

    Strengths

    • Ranks #4 of 22 compared models
    • 86th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures Winogrande, not total model capability
  5. 05
    NV
    Llama 3.1 Nemotron 70B InstructNVIDIA
    Score
    84.5%

    Strengths

    • Ranks #5 of 22 compared models
    • 81th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures Winogrande, not total model capability

Selection summary

Best AI Models for Winogrande

GPT-4 currently leads Winogrande with 87.5%. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price and runtime.

Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.

Benchmark rank #1GPT-487.5% · $30 input / $60 output per 1M tokensBenchmark rank #2MiMo-V2.5-Pro85.6% · $0.43 input / $0.87 output per 1M tokensBenchmark rank #3Command R+85.4% · $2.5 input / $10 output per 1M tokens

What is Winogrande?

What Winogrande measures and how its scores work.

WinoGrande: An Adversarial Winograd Schema Challenge at Scale. A large-scale dataset of 44,000 pronoun resolution problems designed to test machine commonsense reasoning. Uses adversarial filtering to reduce spurious biases and provides a more robust evaluation of whether AI systems truly understand commonsense or exploit statistical shortcuts. Current best AI methods achieve 59.4-79.1% accuracy, significantly below human performance of 94.0%.

Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.

Family
Winogrande
Modality
text
Primary category
reasoning
Score direction
higher
LLMBoard eligible
Yes
Evaluation key
winogrande|llm-stats-current

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 87.5%.

What does Winogrande measure?

WinoGrande: An Adversarial Winograd Schema Challenge at Scale. A large-scale dataset of 44,000 pronoun resolution problems designed to test machine commonsense reasoning. Uses adversarial filtering to reduce spurious biases and provides a more robust evaluation of whether AI systems truly understand commonsense or exploit statistical shortcuts. Current best AI methods achieve 59.4-79.1% accuracy, significantly below human performance of 94.0%.

Is a higher score better?

Yes. Higher values rank better for this benchmark.

How many models are compared?

22 model results are currently shown.

Does this benchmark affect the overall score?

Yes. This benchmark can contribute to the current LLMBoard capability score.