MMLU-Pro S
AI benchmark · Knowledge
What is MMLU-Pro?
MMLU-Pro measures language models' knowledge and reasoning through over 12,000 ten-option multiple-choice questions across 14 academic subjects.
General capability
“MMLU-Pro emphasizes deeper reasoning over knowledge recall, creating a more challenging evaluation that better discriminates between advanced language models.”
artificialanalysis.ai/evaluations/mmlu-pro ↗“even the leader clears 95% on math and biology but drops to roughly 87% on law, history, and health”
vals.ai/benchmarks/mmlu_pro ↗“even the leader clears 95% on math and biology but drops to roughly 87% on law, history, and health”
vals.ai/benchmarks/mmlu_pro ↗“even the leader clears 95% on math and biology but drops to roughly 87% on law, history, and health”
vals.ai/benchmarks/mmlu_pro ↗“tests models on multiple-choice questions on subjects like physics, chemistry, etc.”
vals.ai/benchmarks/mmlu_pro ↗
Paper: arxiv.org/abs/2406.01574 ↗
Lineage
Benchmarks this one is built on and built from: versions, subsets, extensions and the indexes that combine them.
Built on
MMLU original
Derived benchmarks
MMLU-ProX extension
Included in
“MMLU-Pro, GPQA, AIME, HLE, ARC-AGI-2, Metacog, SWE-Pro, IFEval, LCB, **Union Eval**, etc.”
huggingface.co/datasets/youssef3146/ALL-Bench-Leaderboard ↗
Tier S
The standard: frontier labs report it, or a trusted evaluator scores hundreds of models on it.
How much stands behind a tier: usage, who backs it, a confirmed line from its own site or paper, and a clear read on X. High with three of the four, medium with two, low otherwise.
- 313 models scored
- in curated RL directories (2 pts)
On X · well regarded, moderate buzz, from 9 posts about MMLU-Pro in the last month
MMLU-Pro is routinely cited as a standard quality check in quantization and local-model evaluations, though one preprint frames it as a response to saturation.
How we got to tier S
- Adoption: 313 models scored. Tier S.
- On X: well regarded, moderate buzz. “MMLU-Pro is routinely cited as a standard quality check in quantization and local-model evaluations, though one preprint frames it as a response to saturation.” No move.
- In curated RL directories: 2 points, a floor of C.
- Confirmed: a line from its own site or paper places it.
A clear view on X moves a tier one step up or down; most entries don't move. How tiers work.
Where MMLU-Pro sits
Its domain on the map, live. Hover a logo for its name; click to open it. Open it on the full map
Next: browse all benchmarks. Entry from the RL Research daily scrape of public sources, 2026-11-05 snapshot.