Gemini 3.5 Flash Lite (Medium): benchmark results

Provider: Google. Released 2026-07-21. Access: API.

Unified ELO 1568 ± 1, rank #576 of 2032 rated models, from 19 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
DelistBench (Closed-Book)66.5Overall joint accuracy within 7 days (%; all 1,200 security-75
DelistBench - Announcement Status (Closed-Book)85.4Announcement-status accuracy (%; whether each of 1,200 secur75
ParseBench - Visual Grounding65.55Element Pass Rate68.8
Chinese Competitive Debating - Winner Prediction59.5Accuracy (%; the model reads a full match transcript and giv43.8
Context Arena43.57Average Score (%)39.5
ARC-AGI-25.28Accuracy (%)39.3
ParseBench54.3Overall Score29.5
ARC-AGI-132.33Accuracy (%)25.9
ParseBench - Tables72.85GTRM Score (GriTS + TableRecordMatch)25.7
ParseBench - Semantic Formatting48.05Semantic Formatting Score20
Maze-Bench15.29Mean Score19.4
ParseBench - Charts6.23ChartDataPointMatch Score17.1

Interactive version: theaggregate.ai/model?slug=gemini-3-5-flash-lite-medium · How It Works · Data refreshed daily, snapshot 2026-09-26.