Gemini 2.5 Flash Lite is a lightweight reasoning model in the Gemini 2.5 family, optimized for ultra-low latency and cost efficiency. It delivers faster token generation and improved benchmark performance compared to earlier Flash models, with thinking disabled by default to prioritize speed. Designed for high-throughput use cases where rapid response is more important than deep reasoning, it offers the most affordable entry point in the Gemini 2.5 lineup.
API|VisionReasoningWeb SearchFile|Proprietary Model
Knowledge Cutoff
2025-01-31
Input → Output Format
Context Memory
1.0MIN66KOUT
AI Performance Evaluation
Arena Overall Score
1380
±4As of 2026-04-23
Overall Rank
No.126
47,291 Votes
Arena by Ability
Hard Prompts
1390±5No.133
Expert Knowledge
1386±12No.127
Instruction Following
1365±6No.132
Conversation Memory
1374±7No.129
Creative
1361±8No.109
Coding
1397±7No.149
Math
1364±11No.140
Arena by Occupation
Creative Writing
1371±6No.110
Social Sciences
1403±7No.119
Media
1346±7No.124
Business
1378±7No.124
Healthcare
1399±12No.124
Legal
1400±11No.116
Software
1400±5No.141
Mathematics
1370±13No.135
Source:Arena Intelligence
Overall
AA Intelligence Index
19%↓19%
LiveBench
42%↓19%
ForecastBench
57%↓2%
Reasoning & Math
AA Math Index
47%↓27%
GPQA Diamond
65%↓16%
HLE
4.6%↓13%
MMLU-Pro
80%↓2%
AIME 2025
47%↓27%
LB Reasoning
43%↓16%
LB Math
61%↓13%
LB Data
47%↓3%
Coding
AA Coding Index
15%↓20%
LiveCodeBench
64%↓1%
LB Coding
66%↓7%
LB Agentic
5.0%↓38%
TAU2
30%↓43%
TerminalBench
7.6%↓23%
SciCode
28%↓12%
Language & Instructions
IFBench
42%↓15%
AA-LCR
48%↓14%
Hallucination (HHEM)
3.3%↓7%
Factual (HHEM)
97%↑7%
LB Language
52%↓20%
LB IF
23%↓23%
Output Speed
Standard Mode
105tok/s↑23
First Output 0.53s
