Gemma 4 E4B — benchmark results

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

Unified ELO 1292 ± 15, rank #1546 of 1776 rated models, from 312 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Managing Procedural Memory in LLM Agents3size (self-reported)95.8
POLAR-Bench90.87Overall Mean (self-reported)85.7
P3B394.4LLM pt-PT (↑) (self-reported)84.2
BIG-Bench Extra Hard33.1Score (self-reported)73.3
MathVision59.5Overall Accuracy (%)73.3
CritPt0.6Accuracy (self-reported)71.1
SEA-HELM62.57Mean Score (%)69.8
EuroEval Lithuanian NLU - MultiWikiQA LT60.71Reading comprehension Score (%)67.7
EuroEval Bulgarian NLU - Cinexio40.53Sentiment classification Score (%)67.3
EuroEval Ukrainian NLU - MultiWikiQA UK57.72Reading comprehension Score (%)67.2
EuroEval Albanian NLU - MultiWikiQA SQ55.2Reading comprehension Score (%)64.6
AgentFloor46TCR (self-reported)62.5

Interactive version: theaggregate.ai/model?slug=gemma-4-e4b · How the rankings work · Data refreshed daily, snapshot 2026-07-22.