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

MGSM Leaderboard

MGSM (Multilingual Grade School Math) is a benchmark of grade-school math problems. Contains 250 grade-school math problems manually translated from the GSM8K dataset into ten typologically diverse languages: Spanish, French, German, Russian, Chinese, Japanese, Thai, Swahili, Bengali, and Telugu. Evaluates multilingual mathematical reasoning capabilities.

Updated Aug 17, 2026

Models31
Model coverage31
MetricScore
EvidenceB

On this page

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

MGSM Ranking

Higher score ranks better on this benchmark.

30 of 31 rows
Columns

Show columns

Sort by
Rank
Model
Score
Percentile
Participants
Evidence
Evaluated
Rank01ModelMELlama 4 MaverickMetaScore92.3%Percentile100.0%Participants31EvidenceCEvaluatedAug 17, 2026
Rank02ModelOPo3-miniOpenAIScore92.0%Percentile96.7%Participants31EvidenceCEvaluatedAug 17, 2026
Rank03ModelANClaude 3.5 SonnetAnthropicScore91.6%Percentile93.3%Participants31EvidenceCEvaluatedAug 17, 2026
Rank04ModelANClaude 3.5 SonnetAnthropicScore91.6%Percentile90.0%Participants31EvidenceCEvaluatedAug 17, 2026
Rank05ModelMELlama 3.3 70B InstructMetaScore91.1%Percentile86.7%Participants31EvidenceCEvaluatedAug 17, 2026
Rank06ModelOPo1-previewOpenAIScore90.8%Percentile83.3%Participants31EvidenceCEvaluatedAug 17, 2026
Rank07ModelANClaude 3 OpusAnthropicScore90.7%Percentile80.0%Participants31EvidenceCEvaluatedAug 17, 2026
Rank08ModelMELlama 4 ScoutMetaScore90.6%Percentile76.7%Participants31EvidenceCEvaluatedAug 17, 2026
Rank09ModelOPGPT-4oOpenAIScore90.5%Percentile73.3%Participants31EvidenceCEvaluatedAug 17, 2026
Rank10ModelOPo1OpenAIScore89.3%Percentile70.0%Participants31EvidenceCEvaluatedAug 17, 2026
Rank11ModelOPGPT-4 TurboOpenAIScore88.5%Percentile66.7%Participants31EvidenceCEvaluatedAug 17, 2026
Rank12ModelGOGemini 1.5 ProGoogleScore87.5%Percentile63.3%Participants31EvidenceCEvaluatedAug 17, 2026
Rank13ModelOPGPT-4o miniOpenAIScore87.0%Percentile60.0%Participants31EvidenceCEvaluatedAug 17, 2026
Rank14ModelMELlama 3.2 90B InstructMetaScore86.9%Percentile56.7%Participants31EvidenceCEvaluatedAug 17, 2026
Rank15ModelANClaude 3.5 HaikuAnthropicScore85.6%Percentile53.3%Participants31EvidenceCEvaluatedAug 17, 2026
Rank16ModelACQwen3 235B A22BAlibaba Cloud / Qwen TeamScore83.5%Percentile50.0%Participants31EvidenceCEvaluatedAug 17, 2026
Rank17ModelANClaude 3 SonnetAnthropicScore83.5%Percentile46.7%Participants31EvidenceCEvaluatedAug 17, 2026
Rank18ModelGOGemini 1.5 FlashGoogleScore82.6%Percentile43.3%Participants31EvidenceCEvaluatedAug 17, 2026
Rank19ModelMIPhi 4MicrosoftScore80.6%Percentile40.0%Participants31EvidenceCEvaluatedAug 17, 2026
Rank20ModelANClaude 3 HaikuAnthropicScore75.1%Percentile36.7%Participants31EvidenceCEvaluatedAug 17, 2026
Rank21ModelOPGPT-4OpenAIScore74.5%Percentile33.3%Participants31EvidenceCEvaluatedAug 17, 2026
Rank22ModelMELlama 3.2 11B InstructMetaScore68.9%Percentile30.0%Participants31EvidenceCEvaluatedAug 17, 2026
Rank23ModelGOGemma 3n E4B InstructedGoogleScore67.0%Percentile26.7%Participants31EvidenceCEvaluatedAug 17, 2026
Rank24ModelMIPhi 4 MiniMicrosoftScore63.9%Percentile23.3%Participants31EvidenceCEvaluatedAug 17, 2026
Rank25ModelGOGemma 3n E4B Instructed LiteRT PreviewGoogleScore60.7%Percentile20.0%Participants31EvidenceCEvaluatedAug 17, 2026
Rank26ModelMIPhi-3.5-MoE-instructMicrosoftScore58.7%Percentile16.7%Participants31EvidenceCEvaluatedAug 17, 2026
Rank27ModelMELlama 3.2 3B InstructMetaScore58.2%Percentile13.3%Participants31EvidenceCEvaluatedAug 17, 2026
Rank28ModelOPGPT-3.5 TurboOpenAIScore56.3%Percentile10.0%Participants31EvidenceBEvaluatedAug 17, 2026
Rank29ModelGOGemma 3n E2B InstructedGoogleScore53.1%Percentile6.7%Participants31EvidenceCEvaluatedAug 17, 2026
Rank30ModelGOGemma 3n E2B Instructed LiteRT (Preview)GoogleScore53.1%Percentile3.3%Participants31EvidenceCEvaluatedAug 17, 2026
Rank31ModelMIPhi-3.5-mini-instructMicrosoftScore47.9%Percentile0.0%Participants31EvidenceCEvaluatedAug 17, 2026

MGSM Highlights

The leading models and scores on this benchmark.

Rank #1Llama 4 Maverick92.3%Rank #2o3-mini92.0%Rank #3Claude 3.5 Sonnet91.6%Rank #4Claude 3.5 Sonnet91.6%

MGSM Score Distribution

A closer view of the leading scores on this benchmark.

MGSM

The Top AI Models for MGSM

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

Ranking basisThis mgsm 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
    ME
    Llama 4 MaverickMeta
    Score
    92.3%
    Speed
    Up to 639 tok/s via Sambanova

    Strengths

    • Ranks #1 of 31 compared models
    • 100th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures MGSM, not total model capability
  2. 02
    OP
    o3-miniOpenAI
    Score
    92.0%
    Price
    $1.1 input / $4.4 output per 1M tokens
    Speed
    Up to 115 tok/s via Azure

    Strengths

    • Ranks #2 of 31 compared models
    • 97th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures MGSM, not total model capability
  3. 03
    AN
    Claude 3.5 SonnetAnthropic
    Score
    91.6%
    Speed
    Up to 101 tok/s via Bedrock

    Strengths

    • Ranks #3 of 31 compared models
    • 93th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures MGSM, not total model capability
  4. 04
    AN
    Claude 3.5 SonnetAnthropic
    Score
    91.6%
    Speed
    Up to 101 tok/s via Bedrock

    Strengths

    • Ranks #4 of 31 compared models
    • 90th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures MGSM, not total model capability
  5. 05
    ME
    Llama 3.3 70B InstructMeta
    Score
    91.1%
    Speed
    Up to 2,220 tok/s via Cerebras

    Strengths

    • Ranks #5 of 31 compared models
    • 87th percentile on this benchmark
    • C evidence result

    Considerations

    • This result measures MGSM, not total model capability

Selection summary

Best AI Models for MGSM

Llama 4 Maverick currently leads MGSM with 92.3%. 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 #1Llama 4 Maverick92.3% · Up to 639 tok/s via SambanovaBenchmark rank #2o3-mini92.0% · $1.1 input / $4.4 output per 1M tokensBenchmark rank #3Claude 3.5 Sonnet91.6% · Up to 101 tok/s via Bedrock

What is MGSM?

What MGSM measures and how its scores work.

MGSM (Multilingual Grade School Math) is a benchmark of grade-school math problems. Contains 250 grade-school math problems manually translated from the GSM8K dataset into ten typologically diverse languages: Spanish, French, German, Russian, Chinese, Japanese, Thai, Swahili, Bengali, and Telugu. Evaluates multilingual mathematical reasoning capabilities.

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

Family
MGSM
Modality
text
Primary category
math
Score direction
higher
LLMBoard eligible
Yes
Evaluation key
mgsm|llm-stats-current

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

FAQ

Common questions about MGSM.

Which model scores highest on MGSM?

Llama 4 Maverick is currently ranked first with 92.3%.

What does MGSM measure?

MGSM (Multilingual Grade School Math) is a benchmark of grade-school math problems. Contains 250 grade-school math problems manually translated from the GSM8K dataset into ten typologically diverse languages: Spanish, French, German, Russian, Chinese, Japanese, Thai, Swahili, Bengali, and Telugu. Evaluates multilingual mathematical reasoning capabilities.

Is a higher score better?

Yes. Higher values rank better for this benchmark.

How many models are compared?

31 model results are currently shown.

Does this benchmark affect the overall score?

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