Gemma 4 E4B (Non-reasoning): benchmark results
Provider: Google. Released 2026-04-02. Access: Open.
Aggregate Rating 1384 ± 23, rank #1535 of 2170 rated models, from 32 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MEDit-Bench | 16.9 | mIoU (%; temporal IoU between the union of the predicted cut | 50 |
| AA CritPt | 0.29 | Accuracy (%) | 45.9 |
| AA Terminal-Bench Hard | 7.58 | Accuracy (%) | 40 |
| AA IFBench | 40.54 | Accuracy (%) | 39.6 |
| Tau2-Bench Telecom | 26.02 | Success Rate (%) | 31.2 |
| AA Omniscience - Software Engineering (SWE) - Julia | 4 | Accuracy (%) | 30.4 |
| AA TAU-2 Bench | 26.02 | Accuracy (%) | 30.3 |
| AA GPQA Diamond | 54.95 | Accuracy (%) | 28.8 |
| MEDit-Bench - F1@0.5 | 7.1 | F1@0.5 (%; cut-level F1 after one-to-one matching of predict | 25 |
| AA Humanity's Last Exam | 4.82 | Accuracy (%) | 24.6 |
| AA Long Context Reasoning | 24 | Accuracy (%) | 24.5 |
| AA Omniscience - Software Engineering (SWE) - Rust | 36 | Accuracy (%) | 21.7 |
Interactive version: theaggregate.ai/model?slug=gemma-4-e4b-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-05.