MMLU-Pro
A more robust and challenging successor to MMLU with over 12,000 reasoning-focused questions across 14 subjects, expanding answer choices from four to ten to better discriminate frontier large language models.
What this benchmark measures
A more robust and challenging successor to MMLU with over 12,000 reasoning-focused questions across 14 subjects, expanding answer choices from four to ten to better discriminate frontier large language models.
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 MMLU-Pro, not compared as part of a site-wide ranking.
Frequently asked
What is MMLU-Pro?
A more robust and challenging successor to MMLU with over 12,000 reasoning-focused questions across 14 subjects, expanding answer choices from four to ten to better discriminate frontier large language models. It is a reasoning benchmark measured by accuracy.
What does accuracy mean on MMLU-Pro?
MMLU-Pro reports accuracy (%); higher is better. Scores are shown only within MMLU-Pro and are never averaged with other benchmarks.
What is the top reported MMLU-Pro score?
Gemini 3.1 Pro Preview has the top reported score on MMLU-Pro: 90.99% (accuracy).
Why do MMLU-Pro 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. MMLU-Pro scores are shown within their own metric; evals.report never combines benchmarks into a composite ranking or a single "best model".