Gemma 4 12B: benchmark results

Provider: Google. Released 2026-04-02. Access: Open.

Unified ELO 1582 ± 1, rank #314 of 1392 rated models, from 24 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AI for Education Pedagogy - Technology85.85Accuracy (%)89.5
SEA-HELM65.23Mean Score (%)81.4
BenchmarkList ECI120.45Capability Index (ECI)63.3
AI for Education SEND77.06Accuracy (%)59.5
AI for Education Pedagogy - Primary85.45Accuracy (%)59.2
AI for Education Pedagogy - Secondary81.29Accuracy (%)58.8
ZeroEval GPQA Diamond78.8GPQA Diamond Score58.3
AI for Education Pedagogy81.65Accuracy (%)57.8
AI for Education Pedagogy - Maths80.16Accuracy (%)56.9
AI for Education Pedagogy - Science82.51Accuracy (%)56.3
LLM Stats (MathVision)79.7Score (%)55.9
AI for Education Pedagogy - Social studies78.18Accuracy (%)53.6

Interactive version: theaggregate.ai/model?slug=gemma-4-12b · How It Works · Data refreshed daily, snapshot 2026-09-05.