0

GPT-5

Closed

OpenAI's August 2025 unified frontier model that auto-routes between a fast model and a deeper "thinking" variant.

Open
Publisher
OpenAI
Family
GPT 5
Params
Undisclosed (closed)
Context
400K
Input
Output
Providers
openai
License
Proprietary
Released
7 Aug 2025
Canonical
openai.com/index/introducing-gpt-5

Cite

Notes

Only stored in your browser.

Attribution

Benchmark scores
AiderAAprime-hubVaultdabstep-orgfrontiermathSWE-benchVals AI
Attribution policy →
Context
400K
Input
$1.25
/1M tok
Output
$10
/1M tok
Blended
$3.44
#362/ 414/1M · 3:1
Speed
176
#39/ 414tok/s
Cache read
$0.13
0.10x input

Price & speed via Artificial Analysis · ranked across tracked models · caching via OpenRouter

Reported on 60 evals across 14 domains - best 100.0% on Gutenberg Env · #1 of 5

Reported eval scores

60
Chart Extraction4.6
Mastermind1.6
Gutenberg Env100
Mini Swe Agent Bench93.3
MCP Universe92.7
MCP Universe92.7
Fingpt Sentiment91.7
Github MCP88.9
Transformerpuzzles87.5
Deepscaler80
Deepscaler Math80
Capa68.1
Multi Challenge60
Vpct 154.8
Hypothesis Forge48.9
Arc Agi 30
MCP Env0
Wiki Search90
Wiki Search90
Twenty Questions0.9
Infraresolutionbench83.4
Pii Masking0
LLM Training Puzzles37.5
Acereason Math46.7
Mastermind0.5
Banking Mrm0.2
Mlebench10
Mnist Adversarial0
Aider Polyglot Benchmark88
SWE-bench Verified74.4
LiveCodeBench54.3
SWE-bench74.9
MATH100
AIME 2024: Problems from the American Invitational Mathematics Examination94.6
FrontierMath32.4
MATH-50086.1
AgentHarm: Harmfulness Potential in AI Agents91.1
MMLU-Pro82
Humanity's Last Exam (HLE)5.8
DABstep57.7
BrowseComp0
τ²-bench (Tau²-bench)0
MedScribe83.7
MedCode49.6
SciCode37.8
GPQA Diamond68.6
GPQA (Full Set)83.3
TaxEval v273.4
CorpFin v261.1
MortgageTax65.5
IFBench45
Terminal-Bench (Hard)12.9
AIME 2025: Problems from the American Invitational Mathematics Examination48.3
ProofBench18
SAGE43.7
LegalBench58.2
Vibe Code Bench v1.120.1

Introduced in

paperIntroducing GPT-5