Gemini 3.1 Flash Lite (Non-reasoning) Benchmark Results

Gemini 3.1 Flash Lite (Non-reasoning): benchmark results

Provider: Google. Released 2026-05-07. Access: API.

Unified ELO 1585 ± 23, rank #708 of 2055 rated models, from 18 benchmark results.

Benchmark	Score	Metric	Percentile
UAV-DualCog - Landmark Visibility Counting	59.9	Visibility-count accuracy (%; 1,024 flight videos in which t	96.4
UAV-DualCog - Self-Relative Position	33.7	Answer accuracy (%; 1,024 questions asking where the UAV is	80
UAV-DualCog - Flight Behavior Recognition (Composite)	22.8	Composite-level behavior accuracy (%; 1,024 first-person UAV	65.4
UAV-DualCog - Future Observation Prediction	30.1	Answer accuracy (%; 1,024 questions asking which view the UA	61.8
UAV-DualCog - Landmark-Relative Direction	47.6	Answer accuracy (%; 1,024 questions asking where a landmark	55.7
Hy-MultiTurn - Action Suppression	56.2	Importance-weighted score (%) on the 35 action suppression d	47.6
Hy-MultiTurn - Constraint Memory	34.4	Importance-weighted score (%) on the 34 constraint memory di	28.6
UAV-DualCog - Landmark-Driven Action Decision	35.9	Answer accuracy (%; 1,024 questions asking which way the UAV	20
UAV-DualCog - Flight Behavior Recognition (Atomic)	23.3	Atomic-level behavior accuracy (%; the atomic flight actions	19.2
LLM Milgram Obedience - Baseline Full-Obedience Rate	100	Valid baseline sessions ending at 450 V (%; lower is less ob	5.3
LLM Milgram Obedience - Baseline Mean Breakoff Voltage	450	Mean last shock administered in valid baseline sessions (vol	5.3
Hy-MultiTurn	38.9	Importance-weighted score (%): share of the importance-weigh	0