ModelBeat
Models/GPT-5 mini
All models
O

OpenAI: GPT-5 mini

openai/gpt-5-mini

GPT-5 Mini is a compact version of GPT-5, designed to handle lighter-weight reasoning tasks.

Context
400K
Input
$0.25 / 1M
Output
$2 / 1M
Released
Aug 7, 2025
Modalities
textimagefiletext

Benchmarks

Scores on standardized evaluations. Higher is better — and rank shows where GPT-5 mini lands among all models tracked on Model Beat.

31.0
Intelligence Index
Epoch AI
31st percentile of tracked models
17.0
Coding Index
Epoch AI
17th percentile of tracked models
28.0
Agentic Index
Epoch AI
28th percentile of tracked models

Reasoning

6 evals
ARC-AGI54.3%

Abstract visual reasoning

ARC-AGI-24.4%

Harder abstract reasoning

GPQA Diamond75.0%

Graduate-level scientific reasoning

Humanity's Last Exam19.4%

Frontier of human expert knowledge

SimpleQA Verified21.0%

Factual accuracy & hallucination

WeirdML52.7%

Novel ML problem-solving

Coding

1 evals
SWE-bench Verified64.7%

Real GitHub issue resolution

Math

4 evals
AIME 2024/202586.7%

Olympiad-qualifier math

FrontierMath27.2%

Research-level math problems

FrontierMath Tier 46.3%

Hardest research math

MATH Level 597.8%

Hardest competition math

Agentic & Tools

1 evals
Terminal-Bench34.8%

Command-line agentic tasks

In the news

1

Model & benchmark data from Epoch AI (CC BY); pricing, specs & descriptions from OpenRouter.