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

math benchmark

MMLU-Redux Leaderboard

An improved version of the MMLU benchmark featuring manually re-annotated questions to identify and correct errors in the original dataset. Provides more reliable evaluation metrics for language models by addressing dataset quality issues found in the original MMLU.

Updated Aug 17, 2026

Models48
Model coverage48
MetricScore
EvidenceB

On this page

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

MMLU-Redux Ranking

Higher score ranks better on this benchmark.

30 of 48 rows
Columns

Show columns

Sort by
Rank
Model
Score
Percentile
Participants
Evidence
Evaluated
Rank01ModelACQwen3.7 MaxAlibaba Cloud / Qwen TeamScore95.0%Percentile100.0%Participants48EvidenceCEvaluatedAug 17, 2026
Rank02ModelACQwen3.5-397B-A17BAlibaba Cloud / Qwen TeamScore94.9%Percentile97.9%Participants48EvidenceCEvaluatedAug 17, 2026
Rank03ModelACQwen3.6 PlusAlibaba Cloud / Qwen TeamScore94.5%Percentile95.7%Participants48EvidenceCEvaluatedAug 17, 2026
Rank04ModelACQwen3.7-PlusAlibaba Cloud / Qwen TeamScore94.5%Percentile93.6%Participants48EvidenceCEvaluatedAug 17, 2026
Rank05ModelMAKimi K2-Thinking-0905Moonshot AIScore94.4%Percentile91.5%Participants48EvidenceCEvaluatedAug 17, 2026
Rank06ModelACQwen3.5-122B-A10BAlibaba Cloud / Qwen TeamScore94.0%Percentile89.4%Participants48EvidenceCEvaluatedAug 17, 2026
Rank07ModelACQwen3-235B-A22B-Thinking-2507Alibaba Cloud / Qwen TeamScore93.8%Percentile87.2%Participants48EvidenceCEvaluatedAug 17, 2026
Rank08ModelACQwen3 VL 235B A22B ThinkingAlibaba Cloud / Qwen TeamScore93.7%Percentile85.1%Participants48EvidenceCEvaluatedAug 17, 2026
Rank09ModelACQwen3.6-27BAlibaba Cloud / Qwen TeamScore93.5%Percentile83.0%Participants48EvidenceCEvaluatedAug 17, 2026
Rank10ModelDEDeepSeek-R1-0528DeepSeekScore93.4%Percentile80.8%Participants48EvidenceCEvaluatedAug 17, 2026
Rank11ModelACQwen3.5-35B-A3BAlibaba Cloud / Qwen TeamScore93.3%Percentile78.7%Participants48EvidenceCEvaluatedAug 17, 2026
Rank12ModelACQwen3.6-35B-A3BAlibaba Cloud / Qwen TeamScore93.3%Percentile76.6%Participants48EvidenceCEvaluatedAug 17, 2026
Rank13ModelACQwen3.5-27BAlibaba Cloud / Qwen TeamScore93.2%Percentile74.5%Participants48EvidenceCEvaluatedAug 17, 2026
Rank14ModelACQwen3-235B-A22B-Instruct-2507Alibaba Cloud / Qwen TeamScore93.1%Percentile72.3%Participants48EvidenceCEvaluatedAug 17, 2026
Rank15ModelXIMiMo-V2.5-ProXiaomiScore92.8%Percentile70.2%Participants48EvidenceCEvaluatedAug 17, 2026
Rank16ModelMAKimi K2 InstructMoonshot AIScore92.7%Percentile68.1%Participants48EvidenceCEvaluatedAug 17, 2026
Rank17ModelMAKimi K2-Instruct-0905Moonshot AIScore92.7%Percentile66.0%Participants48EvidenceCEvaluatedAug 17, 2026
Rank18ModelACQwen3-Next-80B-A3B-ThinkingAlibaba Cloud / Qwen TeamScore92.5%Percentile63.8%Participants48EvidenceCEvaluatedAug 17, 2026
Rank19ModelACQwen3 VL 235B A22B InstructAlibaba Cloud / Qwen TeamScore92.2%Percentile61.7%Participants48EvidenceCEvaluatedAug 17, 2026
Rank20ModelACQwen3 VL 32B ThinkingAlibaba Cloud / Qwen TeamScore91.9%Percentile59.6%Participants48EvidenceCEvaluatedAug 17, 2026
Rank21ModelDEDeepSeek-V3.1DeepSeekScore91.8%Percentile57.5%Participants48EvidenceCEvaluatedAug 17, 2026
Rank22ModelACQwen3.5-9BAlibaba Cloud / Qwen TeamScore91.1%Percentile55.3%Participants48EvidenceCEvaluatedAug 17, 2026
Rank23ModelACQwen3-Next-80B-A3B-InstructAlibaba Cloud / Qwen TeamScore90.9%Percentile53.2%Participants48EvidenceCEvaluatedAug 17, 2026
Rank24ModelACQwen3 VL 30B A3B ThinkingAlibaba Cloud / Qwen TeamScore90.9%Percentile51.1%Participants48EvidenceCEvaluatedAug 17, 2026
Rank25ModelACQwen3 VL 32B InstructAlibaba Cloud / Qwen TeamScore89.8%Percentile48.9%Participants48EvidenceCEvaluatedAug 17, 2026
Rank26ModelMELongCat-Flash-ThinkingMeituanScore89.3%Percentile46.8%Participants48EvidenceCEvaluatedAug 17, 2026
Rank27ModelDEDeepSeek-V3DeepSeekScore89.1%Percentile44.7%Participants48EvidenceCEvaluatedAug 17, 2026
Rank28ModelACQwen3 VL 8B ThinkingAlibaba Cloud / Qwen TeamScore88.8%Percentile42.5%Participants48EvidenceCEvaluatedAug 17, 2026
Rank29ModelACQwen3.5-4BAlibaba Cloud / Qwen TeamScore88.8%Percentile40.4%Participants48EvidenceCEvaluatedAug 17, 2026
Rank30ModelACQwen3 VL 30B A3B InstructAlibaba Cloud / Qwen TeamScore88.4%Percentile38.3%Participants48EvidenceCEvaluatedAug 17, 2026
Rank31ModelACQwen3 235B A22BAlibaba Cloud / Qwen TeamScore87.4%Percentile36.2%Participants48EvidenceCEvaluatedAug 17, 2026
Rank32ModelACQwen2.5 72B InstructAlibaba Cloud / Qwen TeamScore86.8%Percentile34.0%Participants48EvidenceCEvaluatedAug 17, 2026
Rank33ModelACQwen3 VL 4B ThinkingAlibaba Cloud / Qwen TeamScore86.0%Percentile31.9%Participants48EvidenceCEvaluatedAug 17, 2026
Rank34ModelACQwen3 VL 8B InstructAlibaba Cloud / Qwen TeamScore84.9%Percentile29.8%Participants48EvidenceCEvaluatedAug 17, 2026
Rank35ModelACQwen2.5 32B InstructAlibaba Cloud / Qwen TeamScore83.9%Percentile27.7%Participants48EvidenceCEvaluatedAug 17, 2026
Rank36ModelMAMinistral 3 (14B Base 2512)Mistral AIScore82.0%Percentile25.5%Participants48EvidenceCEvaluatedAug 17, 2026
Rank37ModelMAMistral Large 3Mistral AIScore82.0%Percentile23.4%Participants48EvidenceCEvaluatedAug 17, 2026
Rank38ModelACQwen3 VL 4B InstructAlibaba Cloud / Qwen TeamScore81.5%Percentile21.3%Participants48EvidenceCEvaluatedAug 17, 2026
Rank39ModelACQwen2.5 14B InstructAlibaba Cloud / Qwen TeamScore80.0%Percentile19.1%Participants48EvidenceCEvaluatedAug 17, 2026
Rank40ModelACQwen3.5-2BAlibaba Cloud / Qwen TeamScore79.6%Percentile17.0%Participants48EvidenceCEvaluatedAug 17, 2026
Rank41ModelMAMinistral 3 (8B Base 2512)Mistral AIScore79.3%Percentile14.9%Participants48EvidenceCEvaluatedAug 17, 2026
Rank42ModelACQwen2.5-Coder 32B InstructAlibaba Cloud / Qwen TeamScore77.5%Percentile12.8%Participants48EvidenceCEvaluatedAug 17, 2026
Rank43ModelACQwen2.5 7B InstructAlibaba Cloud / Qwen TeamScore75.4%Percentile10.6%Participants48EvidenceCEvaluatedAug 17, 2026
Rank44ModelMAMinistral 3 (3B Base 2512)Mistral AIScore73.5%Percentile8.5%Participants48EvidenceCEvaluatedAug 17, 2026
Rank45ModelACQwen2.5-Omni-7BAlibaba Cloud / Qwen TeamScore71.0%Percentile6.4%Participants48EvidenceCEvaluatedAug 17, 2026
Rank46ModelACQwen2.5-Coder 7B InstructAlibaba Cloud / Qwen TeamScore66.6%Percentile4.3%Participants48EvidenceCEvaluatedAug 17, 2026
Rank47ModelACQwen3.5-0.8BAlibaba Cloud / Qwen TeamScore59.5%Percentile2.1%Participants48EvidenceCEvaluatedAug 17, 2026
Rank48ModelBAERNIE 4.5BaiduScore43.2%Percentile0.0%Participants48EvidenceCEvaluatedAug 17, 2026

MMLU-Redux Highlights

The leading models and scores on this benchmark.

Rank #1Qwen3.7 Max95.0%Rank #2Qwen3.5-397B-A17B94.9%Rank #3Qwen3.6 Plus94.5%Rank #4Qwen3.7-Plus94.5%

MMLU-Redux Score Distribution

A closer view of the leading scores on this benchmark.

MMLU-Redux

The Top AI Models for MMLU-Redux

The first five results on this benchmark, with official price and output speed added where the model identity can be matched.

Ranking basisThis mmlu-redux 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
    AC
    Qwen3.7 MaxAlibaba Cloud / Qwen Team
    Score
    95.0%
    Price
    $2.5 input / $7.5 output per 1M tokens
    Speed
    Up to 5.8 tok/s via Together

    Strengths

    • Ranks #1 of 48 compared models
    • 100th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures MMLU-Redux, not total model capability
  2. 02
    AC
    Qwen3.5-397B-A17BAlibaba Cloud / Qwen Team
    Score
    94.9%
    Price
    $0.60 input / $3.6 output per 1M tokens

    Strengths

    • Ranks #2 of 48 compared models
    • 98th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures MMLU-Redux, not total model capability
  3. 03
    AC
    Qwen3.6 PlusAlibaba Cloud / Qwen Team
    Score
    94.5%
    Price
    $0.50 input / $3.0 output per 1M tokens
    Speed
    Up to 16 tok/s via Together

    Strengths

    • Ranks #3 of 48 compared models
    • 96th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures MMLU-Redux, not total model capability
  4. 04
    AC
    Qwen3.7-PlusAlibaba Cloud / Qwen Team
    Score
    94.5%
    Price
    $0.50 input / $3.0 output per 1M tokens

    Strengths

    • Ranks #4 of 48 compared models
    • 94th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures MMLU-Redux, not total model capability
  5. 05
    MA
    Kimi K2-Thinking-0905Moonshot AI
    Score
    94.4%
    Price
    $0.60 input / $2.5 output per 1M tokens

    Strengths

    • Ranks #5 of 48 compared models
    • 91th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures MMLU-Redux, not total model capability

Selection summary

Best AI Models for MMLU-Redux

Qwen3.7 Max currently leads MMLU-Redux with 95.0%. 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 #1Qwen3.7 Max95.0% · $2.5 input / $7.5 output per 1M tokensBenchmark rank #2Qwen3.5-397B-A17B94.9% · $0.60 input / $3.6 output per 1M tokensBenchmark rank #3Qwen3.6 Plus94.5% · $0.50 input / $3.0 output per 1M tokens

What is MMLU-Redux?

What MMLU-Redux measures and how its scores work.

An improved version of the MMLU benchmark featuring manually re-annotated questions to identify and correct errors in the original dataset. Provides more reliable evaluation metrics for language models by addressing dataset quality issues found in the original MMLU.

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

Family
MMLU-Redux
Modality
text
Primary category
math
Score direction
higher
LLMBoard eligible
Yes
Evaluation key
mmlu-redux|llm-stats-current

Benchmark scores retain their original unit. Overall score eligibility is shown separately.

FAQ

Common questions about MMLU-Redux.

Which model scores highest on MMLU-Redux?

Qwen3.7 Max is currently ranked first with 95.0%.

What does MMLU-Redux measure?

An improved version of the MMLU benchmark featuring manually re-annotated questions to identify and correct errors in the original dataset. Provides more reliable evaluation metrics for language models by addressing dataset quality issues found in the original MMLU.

Is a higher score better?

Yes. Higher values rank better for this benchmark.

How many models are compared?

48 model results are currently shown.

Does this benchmark affect the overall score?

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