GPT-5 Nano is the smallest and fastest variant in the GPT-5 family, optimized for ultra-low latency and cost-sensitive environments. As the successor to GPT-4.1 Nano, it is purpose-built for developer tools, classification tasks, rapid interactions, and real-time applications where response speed takes priority over deep reasoning. Despite its compact size, it retains key instruction-following and safety features from the GPT-5 series.
API|VisionReasoningWeb SearchFile|Proprietary Model
Knowledge Cutoff
2024-05-31
Input → Output Format
Context Memory
400KIN128KOUT
AI Performance Evaluation
Arena Overall Score
1337
±7As of 2026-04-23
Overall Rank
No.171
8,281 Votes
Arena by Ability
Hard Prompts
1354±10No.166
Expert Knowledge
1353±34No.151
Instruction Following
1325±13No.164
Conversation Memory
1323±16No.183
Creative
1248±20No.230
Coding
1381±15No.169
Math
1344±27No.157
Arena by Occupation
Creative Writing
1290±14No.210
Social Sciences
1345±17No.179
Media
1277±16No.200
Business
1336±16No.168
Healthcare
1359±28No.158
Legal
1330±27No.195
Software
1371±11No.171
Mathematics
1353±29No.153
Source:Arena Intelligence
Overall
AA Intelligence Index
26%↓12%
LiveBench
48%↓12%
ForecastBench
57%↓2%
Reasoning & Math
AA Math Index
78%↑5%
GPQA Diamond
67%↓14%
HLE
7.6%↓10%
MMLU-Pro
77%↓5%
AIME 2025
78%↑5%
LB Reasoning
35%↓24%
LB Math
65%↓9%
LB Data
44%↓5%
Coding
AA Coding Index
23%↓11%
LiveCodeBench
76%↑11%
LB Coding
67%↓6%
LB Agentic
28%↓15%
TAU2
30%↓43%
TerminalBench
17%↓14%
SciCode
34%↓7%
Language & Instructions
IFBench
66%↑9%
AA-LCR
40%↓22%
Hallucination (HHEM)
11%↑0%
Factual (HHEM)
90%↑0%
LB Language
48%↓24%
LB IF
52%↑6%
Output Speed
Standard Mode
135tok/s↑53
First Output 0.73s
Reasoning Mode
164tok/s↑76
First Output 39.73s
