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

What this benchmark measures

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.

Rows on this page are sourced from public benchmark artifacts, leaderboard exports, or source-linked model reports. Each row keeps benchmark version, source model name, and available run details attached to the score.

The metric shown here is accuracy. It should be interpreted within Global-MMLU, not compared as part of a site-wide ranking.

No composite ranking
evals.report never combines benchmarks. accuracy on Global-MMLU is its own number — don’t average it with other metrics.

Frequently asked

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. It is a reasoning benchmark measured by accuracy.

What does accuracy mean on Global-MMLU?

Global-MMLU reports accuracy (%); higher is better. Scores are shown only within Global-MMLU and are never averaged with other benchmarks.

What is the top reported Global-MMLU score?

Gemini 3.1 Pro Preview has the top reported score on Global-MMLU: 93.2% (accuracy).

Why do Global-MMLU scores differ across runs?

Harness, scaffold, reasoning effort, and prompt setup change results, so two runs of the same model can differ. evals.report keeps each score with its run context so the differences stay visible.

Does evals.report rank models across benchmarks?

No. Global-MMLU scores are shown within their own metric; evals.report never combines benchmarks into a composite ranking or a single "best model".