evals.report
BenchmarksLabsCompareRun guidesIn the wild
BenchmarksReasoning

Global-MMLU

A multilingual extension of MMLU covering 42 languages with culturally-sensitive and culturally-agnostic multiple-choice knowledge questions, measuring accuracy across diverse high-, mid-, and low-resource languages.

ReasoningaccuracyHigher is better
ScoresAboutRun this benchmark

What is Global-MMLU?

A multilingual extension of MMLU covering 42 languages with culturally-sensitive and culturally-agnostic multiple-choice knowledge questions, measuring accuracy across diverse high-, mid-, and low-resource languages. evals.report tracks reported Global-MMLU scores with the model, source, status, date, and run caveats attached — official leaderboard scores, vendor-reported launches, and clearly labeled community runs.

Top reported Global-MMLU score: Gemini 3.1 Pro Preview — 93.2% (accuracy).

ModelLabScore↓Source modelStatusDate
Gemini 3.1 Pro PreviewGoogle DeepMind93.2%—UnverifiedFeb 19, 2026Details
Gemini 3 ProGoogle DeepMind92.2%—UnverifiedNov 18, 2025Details
Claude Opus 4.6Anthropic92.2%—UnverifiedFeb 5, 2026Details
Gemini 3 FlashGoogle DeepMind91.4%—UnverifiedDec 17, 2025Details
Claude Opus 4.5Anthropic91.3%—UnverifiedNov 24, 2025Details
GPT-5OpenAI90.7%—UnverifiedAug 7, 2025Details
GPT-5.1OpenAI90.6%—UnverifiedNov 12, 2025Details
Claude Sonnet 4.6Anthropic90.5%—UnverifiedFeb 17, 2026Details
Gemini 2.5 ProGoogle DeepMind90.3%—UnverifiedMar 25, 2025Details
Qwen3.5-397B-A17BOpenAlibaba / Qwen90.0%—UnverifiedFeb 16, 2026Details
GPT-5.2OpenAI89.8%—UnverifiedDec 11, 2025Details
Grok 4xAI89.5%—UnverifiedJul 9, 2025Details
Grok 4.20 beta reasoningxAI89.5%—UnverifiedMar 9, 2026Details
Claude Sonnet 4.5Anthropic89.3%—UnverifiedSep 29, 2025Details
InklingOpenThinking Machines Lab88.7%InklingVerifiedJul 15, 2026Details
GPT-5 miniOpenAI87.4%—UnverifiedAug 7, 2025Details
DeepSeek V3.2OpenDeepSeek86.5%—UnverifiedDec 1, 2025Details
DeepSeek R1OpenDeepSeek86.0%—UnverifiedJan 20, 2025Details
GLM-4.6OpenZ.ai85.6%—UnverifiedSep 30, 2025Details
Grok 4.1 fast reasoningxAI85.6%—UnverifiedNov 19, 2025Details
MiniMax M2.5OpenMiniMax84.2%—UnverifiedFeb 12, 2026Details
Kimi K2.5OpenMoonshot AI84.0%—UnverifiedJan 27, 2026Details
MiniMax M2.1OpenMiniMax84.0%—UnverifiedDec 23, 2025Details
Claude Haiku 4.5Anthropic83.4%—UnverifiedOct 15, 2025Details
Qwen3 MaxAlibaba / Qwen83.3%—UnverifiedSep 5, 2025Details
GPT-OSS-120BOpenOpenAI82.8%—UnverifiedAug 5, 2025Details
DeepSeek V3.1OpenDeepSeek82.7%—UnverifiedAug 21, 2025Details
Llama 4 MaverickOpenMeta82.5%—UnverifiedApr 5, 2025Details
GLM-5OpenZ.ai81.9%—UnverifiedFeb 11, 2026Details
Mistral LargeMistral AI80.3%—UnverifiedFeb 26, 2024Details
GLM-4.7OpenZ.ai79.9%—UnverifiedDec 22, 2025Details
Llama 4 ScoutOpenMeta74.1%—UnverifiedApr 5, 2025Details
Kimi K2 ThinkingOpenMoonshot AI73.5%—UnverifiedNov 6, 2025Details

Each row reports the model’s accuracy on Global-MMLU. Click a row for the full run context.