Gemma 4 E2B (IT) (Non-reasoning): benchmark results

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

Unified ELO 1416 ± 1, rank #2328 of 3078 rated models, from 28 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
FrameBench - Frame Identification - English74Accuracy (%; FrameNet candidate frames)36.8
FrameBench - Frame Identification - Japanese67.8Accuracy (%; Japanese FrameNet candidate frames)32
MERA v2 - NewReasoning24.3Score (%)26.3
MERA v2 - IFHardBench5Score (%)25
MERA v2 - SOB-Hard1.4Score (%)25
MERA v2 - Reasoning22.7Score (%)21.1
MERA v2 - Enantiosemy14.6Score (%)18.4
FrameBench - English68.6Accuracy (%; mean of five prompt templates)15.8
MERA v20.1Total score13.2
MERA v2 - Agentic3.3Score (%)13.2
MERA v2 - Characters29Score (%)13.2
MERA v2 - Humor5.2Score (%)13.2

Interactive version: theaggregate.ai/model?slug=gemma-4-e2b-it-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-19.