OpenAI · GPT-5

GPT-5 Pro

DeprecatedVerified · Aug 28, 2026

The maximum-effort variant of the original GPT-5, now legacy; trades latency for the highest accuracy of the first GPT-5 generation.

profile normalized against the 105-model field

Context window· 400K of 10M4%

Max output· 272K of 384K71%

Output speed· 45 tok/s14%

Affordability· $120.00 / Mtok out33%

Capability breadth· 7 of 1164%

Capability switches · 7 of 11

Reasoning mode

Tool / function use

Streaming

JSON mode

Structured outputs

Prompt caching

Fine-tuning

Web search

Code execution

Vision input

Audio input

This model is deprecated

OpenAI has announced the end of life for this model, effective December 11, 2026. It may be withdrawn from service; treat the specifications and pricing below as subject to removal.

Specifications

Every value carries a primary source and a verification date.

Capacity

Context window

400K

Max output

272K

Pricing

Input $/Mtok

$15.00 / Mtok USD/Mtok

Output $/Mtok

$120.00 / Mtok USD/Mtok

Performance

Output speed

45 tok/s tok/s

Time to first token

1,400 ms ms

Capabilities

Input modalities

Text, image

Output modalities

Text

Reasoning mode

true

Tool / function use

true

Streaming

true

JSON mode

true

Structured outputs

true

Prompt caching

Yes

Code execution

false

Vision input

Yes

API

API model ID

gpt-5-pro

General

Knowledge cutoff

September 30, 2024

Benchmarks

Sourced evaluation scores, each verified against its primary source.

GPQA Diamond

GPQA Diamond: 89.6%

89.6 %Verified

SWE-bench Verified

SWE-bench Verified: 78.4%

78.4 %Verified

AIME 2025

AIME 2025: 96.5%

96.5 %Verified

MMLU-Pro

MMLU-Pro: 88.2%

88.2 %Verified

Humanity's Last Exam (with tools)

Humanity's Last Exam (with tools): 42.1%

42.1 %Verified

LiveCodeBench

LiveCodeBench: 81.3%

81.3 %Verified

Loading…