reasoning benchmark
Global-MMLU-Lite
A lightweight version of Global MMLU benchmark that evaluates language models across multiple languages while addressing cultural and linguistic biases in multilingual evaluation.
Updated Aug 7, 2026
Published models14
Registry coverage14
MetricScore
EvidenceB
Global-MMLU-Lite leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | GO | 89.2% | 100.0% | 14 | C | |
| 2 | GO | 88.6% | 92.3% | 14 | C | |
| 3 | GO | 88.4% | 84.6% | 14 | C | |
| 4 | GO | 81.1% | 76.9% | 14 | C | |
| 5 | GO | 78.2% | 69.2% | 14 | C | |
| 6 | GO | 75.1% | 61.5% | 14 | C | |
| 7 | GO | 69.5% | 53.9% | 14 | C | |
| 8 | GO | 69.1% | 46.1% | 14 | C | |
| 9 | GO | 64.5% | 38.5% | 14 | C | |
| 10 | GO | 64.5% | 30.8% | 14 | C | |
| 11 | GO | 59.0% | 23.1% | 14 | C | |
| 12 | GO | 59.0% | 15.4% | 14 | C | |
| 13 | GO | 54.5% | 7.7% | 14 | C | |
| 14 | GO | 34.2% | 0.0% | 14 | C |
Score distribution
Top published rows on the benchmark's original scale.
Global-MMLU-Lite highlights
The top published results on this benchmark's own scale.
What is Global-MMLU-Lite?
Definition and scoring fields from the benchmark registry.
A lightweight version of Global MMLU benchmark that evaluates language models across multiple languages while addressing cultural and linguistic biases in multilingual evaluation.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- Global-MMLU-Lite
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- global-mmlu-lite|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about Global-MMLU-Lite.
Which model scores highest on Global-MMLU-Lite?
Gemini 2.5 Pro Preview 06-05 is currently ranked first with 89.2%.
What does Global-MMLU-Lite measure?
A lightweight version of Global MMLU benchmark that evaluates language models across multiple languages while addressing cultural and linguistic biases in multilingual evaluation.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
14 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is marked as eligible for the current LLMBoard capability methodology.