Gemma 4: benchmark results
Provider: Google. Access: API.
Unified ELO 1684 ± 1, rank #79 of 1392 rated models, from 15 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| RAI-Bench - Refusal Rate (JD) | 100 | Rate (%) | 97.4 |
| RAI-Bench - Refusal Rate (Career) | 100 | Rate (%) | 93.5 |
| RAI-Bench - Refusal Rate (General) | 82 | Rate (%) | 85.3 |
| Phare - Harm Resistance | 96.51 | Score (%) | 77.1 |
| RAI-Bench - RAG Robustness (LC Abstention) | 87 | Rate (%) | 74.6 |
| RAI-Bench - RAG Robustness (HY Abstention) | 87 | Rate (%) | 73.2 |
| Phare - Jailbreak Resistance | 61.22 | Score (%) | 65.7 |
| RAI-Bench - Refusal Rate (Physics) | 100 | Rate (%) | 63 |
| Phare - Average Safety | 67.39 | Score (%) | 59.1 |
| Phare - Hallucination Resistance | 76.62 | Score (%) | 58.6 |
| ABRA | 68 | Overall Avg (self-reported) | 55.6 |
| FutureEval | 3.33 | Unified Forecasting Score | 40.2 |
Interactive version: theaggregate.ai/model?slug=gemma-4 · How It Works · Data refreshed daily, snapshot 2026-09-05.