Gemini 3 Flash (Medium): benchmark results

Provider: Google. Released 2025-12-17. Access: API.

Aggregate Rating 1735 ± 21, rank #303 of 2287 rated models, from 37 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
CoopEval - Mediation0.87Normalized mean payoff (open scale; below 0 when exploited, 100
GT-HarmBench - Stag Hunt86Utilitarian accuracy (%) on the 317 Stag Hunt scenarios: sha100
PhageBench56.33Accuracy (%) averaged over the five PhageBench tasks (phage 100
PhageBench - Contamination Detection58.83Accuracy (%) on 1,200 two-option items deciding whether a ph100
PhageBench - Host Prediction62.5Accuracy (%) on 1,200 four-option items naming the bacterial100
PhageBench - Lifestyle Classification57.4Accuracy (%) on 1,000 two-option items classifying a phage a100
PhageBench - Phage Contig Identification70.83Accuracy (%) on 1,200 two-option items telling phage contigs100
GT-HarmBench - Chicken92Utilitarian accuracy (%) on the 379 Chicken scenarios: share92.9
ParseBench74.07Overall Score90.1
ParseBench - Tables91.03GTRM Score (GriTS + TableRecordMatch)88.4
CresOWLve - Exact Match35.08Exact-match accuracy (%) after lowercasing and punctuation r82.4
CresOWLve - Russian Exact Match51.82Exact-match accuracy (%) after lowercasing, punctuation remo82.4

Interactive version: theaggregate.ai/model?slug=gemini-3-flash-medium · How It Works · Data refreshed daily, snapshot 2026-11-22.