Gemini 2.5 Flash 04-17 preview — benchmark results

Provider: Google. Released 2025-04-17. Access: API.

Unified ELO 1671 ± 67, rank #383 of 1839 rated models, from 7 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
HELM Safety XSTest98.8LM Evaluated Safety score (%)98.8
HELM Safety BBQ97.7BBQ accuracy (%)91.9
HELM AIR-Bench68.7Refusal Rate (%)58.1
HELM Safety91.2Mean score (self-reported)50.4
HELM Safety Anthropic Red Team98.8LM Evaluated Safety score (%)48.3
HELM Safety SimpleSafetyTests98LM Evaluated Safety score (%)40.1
HELM Safety HarmBench62.6LM Evaluated Safety score (%)26.7

Interactive version: theaggregate.ai/model?slug=gemini-2-5-flash-04-17-preview · How It Works · Data refreshed daily, snapshot 2026-08-05.