Atlas

Benchmarks

← All benchmarks

Global MMLU Lite — Google 15-Language Evaluation

General QA

Google's Gemini 2.0 Flash model card reports Global MMLU Lite over 15 human-translated languages, with 200 culturally sensitive and 200 culturally agnostic questions per language. This 6,000 translated-question aggregate is retained as a diagnostic, separate from Artificial Analysis' 16-language aggregate.

Top models (higher is better)

ModelScore
Gemini 2.0 Flash83.4
Gemini 1.5 Pro 00280.8
Gemini 2.0 Flash-Lite Preview 02-0578.2
Gemini 1.5 Flash 00273.7
Loading Atlas data…