Gemini 2.5 Flash is Google's workhorse reasoning model, designed for fast, high-quality responses across coding, mathematics, and scientific tasks. It features built-in "thinking" capabilities with configurable thinking levels, allowing it to balance response speed and reasoning depth based on task complexity. Supporting a 1M-token context window with multimodal inputs including text, images, audio, video, and PDFs, it delivers strong performance at a fraction of the cost and latency of larger Gemini Pro models.
API|VisionReasoningWeb SearchFile|Proprietary Model
Knowledge Cutoff
2025-01-31
Input → Output Format
Context Memory
1.0MIN66KOUT
AI Performance Evaluation
Arena Overall Score
1411
±3As of 2026-05-01
Overall Rank
No.95
113,223 Votes
Arena by Ability
Hard Prompts
1420±3No.107
Expert Knowledge
1427±8No.95
Instruction Following
1402±4No.88
Conversation Memory
1404±5No.108
Creative
1397±5No.67
Coding
1424±5No.137
Math
1407±7No.95
Arena by Occupation
Creative Writing
1405±4No.72
Social Sciences
1430±5No.92
Media
1389±5No.82
Business
1397±5No.111
Healthcare
1426±8No.106
Legal
1428±7No.79
Software
1422±4No.126
Mathematics
1419±8No.87
Source:Arena Intelligence
Overall
AA Intelligence Index
27%↓12%
LiveBench
47%↓14%
ForecastBench
59%↓1%
Reasoning & Math
AA Math Index
73%↓1%
GPQA Diamond
79%↓3%
HLE
11%↓6%
MMLU-Pro
83%↑2%
AIME 2025
73%↓1%
MATH-500
98%↑5%
LB Reasoning
45%↓24%
LB Math
69%↓5%
LB Data
47%↓6%
Coding
AA Coding Index
22%↓14%
LiveCodeBench
70%↑4%
LB Coding
66%↓7%
LB Agentic
17%↓28%
TAU2
32%↓49%
TerminalBench
14%↓20%
SciCode
39%↓2%
Language & Instructions
IFBench
50%↓13%
AA-LCR
62%↑0%
Hallucination (HHEM)
7.8%↓2%
Factual (HHEM)
92%↑2%
LB Language
62%↓10%
LB IF
29%↓22%
Output Speed
Standard Mode
213tok/s↑135
First Output 0.44s
Reasoning Mode
210tok/s↑123
First Output 14.08s
