Gemini 3.1 Flash Lite

Google's Gemini 3.1 Flash Lite. Optimized for high-volume, cost-efficient reasoning and multimodal workloads.

Reasoning
Vision
Audio Input
Video Input
Native PDF input
Tool Calling
Structured Output

Added Mar 3, 2026

Pricing

Auto routing · per 1M tokens

Input: $0.25
Output: $1.50
Cache read: $0.025

Compare provider prices

Specifications

Context window: 1M
Max output: 65.5K
Avg output (7d): 496 tokens; Longer than 43% of models

Benchmarks

Artificial Analysis

LMArena

Vectara

Sourced from Artificial Analysis.

Intelligence Index

15.6

Better than 57% of models compared

Coding Index

34.7

Better than 41% of models compared

Agentic Index

1.6

Better than 28% of models compared

Agentic work

AutomationBench-AA

Workflow automation with guardrail penalties

6.8%

Better than 31% of models compared

Harvey LAB-AA

Legal agentic work criterion pass rate

31.1%

Better than 2% of models compared

AA-Briefcase

Agentic knowledge work (Elo)

204 Elo

Better than 13% of models compared

GDPval-AA v2

Economically valuable tasks (Elo)

434 Elo

Better than 23% of models compared

Document reasoning

GDP.pdf

Professional PDF reasoning: all-pass rate

7.8%

Better than 34% of models compared

AA-LCR v1.1

Long context reasoning with updated grading

74.3%

Better than 72% of models compared

MLCR-AA

Medical long-context reasoning

6.1%

Better than 19% of models compared

Reasoning

HLE

Humanity's Last Exam

17.2%

Better than 64% of models compared

IFBench

Instruction-following benchmark

77.2%

Better than 96% of models compared

Coding

Terminal-Bench v4.0

Practical coding and terminal tasks

0.5%

Better than 35% of models compared

SciCode

Python programming for scientific computing

43.4%

Better than 35% of models compared

Legacy benchmarks

GPQA Diamond (legacy)

Graduate-level scientific reasoning

82.2%

Better than 70% of models compared

Terminal-Bench Hard (legacy)

Agentic coding and terminal use

24.2%

Better than 64% of models compared

T²-Bench Telecom (legacy)

Conversational AI agents in dual-control scenarios

31.3%

Better than 38% of models compared

AA-LCR (unversioned / legacy)

Long context reasoning evaluation