Atlas

Benchmarks

← All benchmarks

Global MMLU Lite — Google Unspecified Language Cohort

General QA

Global MMLU Lite accuracy reported in Google's Gemini 2.5 Pro, Flash and Flash-Lite model cards, without a verified language roster for these observations. The language count and exact item cohort remain unknown, so these source-reported results are diagnostic and do not enter the Artificial Analysis 16-language factor.

Top models (higher is better)

ModelScore
Gemini 2.5 Pro Experimental 03-2589.8
Gemini 2.5 Pro89.2
Gemini 2.5 Pro Preview 05-0688.6
Gemini 2.5 Flash88.4
Gemini 2.5 Flash Preview 04-1788.4
Gemini 2.5 Flash Preview (09-2025)87.9
Gemini 2.5 Flash-Lite Preview (09-2025)84.9
Gemini 2.5 Flash-Lite84.5
Gemini 2.0 Flash83.4
Loading Atlas data…