GPT-5 (high)

OpenAIGPTProprietary

描述

GPT-5 is a flagship model from OpenAI designed for coding, reasoning, and agentic tasks across domains. It is optimized for coding and agentic tasks with higher reasoning capabilities and medium speed.

發布日期

2025-08-07

參數規模

—

上下文長度

400K

支援模態

image, text

能力雷達圖

general

coding

reasoning

science估算

agents

multimodal

Science 在缺少專門科學評測時使用推理能力代理估算。

排行榜排名

領域	#排名	分數	來源
程式碼能力榜	26	81.0	AA
通用能力榜	35	78.0	AA
數學推理	7	97.0	AA
科學能力	55	70.0	AA

基準測試分數 (LLM Stats)

Agents

BrowseComp

54.9%自報

Biology

GPQA

85.7%自報

Code

SWE-Lancer (IC-Diamond subset)

100.0%自報

HumanEval

93.4%自報

Aider-Polyglot

88.0%自報

SWE-Bench Verified

74.9%自報

Communication

Tau2 Telecom

96.7%自報

Tau2 Retail

81.1%自報

Multi-Challenge

69.6%自報

Tau2 Airline

62.6%自報

Finance

MMLU

92.5%自報

General

MMMU

84.2%自報

MMMU-Pro

78.4%自報

Internal API instruction following (hard)

64.0%自報

LongFact Objects

0.8%自報

LongFact Concepts

0.7%自報

Healthcare

VideoMMMU

84.6%自報

HealthBench Hard

1.6%自報

Language

COLLIE

99.0%自報

Long Context

OpenAI-MRCR: 2 needle 128k

95.2%自報

OpenAI-MRCR: 2 needle 256k

86.8%自報

Math

AIME 2025

94.6%自報

HMMT 2025

93.3%自報

MATH

84.7%自報

FrontierMath

26.3%自報

Humanity's Last Exam

24.8%自報

Multimodal

VideoMME w sub.

86.7%自報

CharXiv-R

81.1%自報

Reasoning

BrowseComp Long Context 128k

90.0%自報

BrowseComp Long Context 256k

88.8%自報

Graphwalks BFS <128k

78.3%自報

Graphwalks parents <128k

73.3%自報

ERQA

65.7%自報

FActScore

1.0%自報

AA 評測指數

Math Index

94.3

Intelligence Index

36.1

Math 500

1.0

Aime

1.0

Aime 25

0.9

Mmlu Pro

0.9

Gpqa

0.9

Tau2

0.8

Livecodebench

0.8

Lcr

0.8

Ifbench

0.7

Scicode

0.4

Terminalbench Hard

0.3

Hle

0.3

LLM Stats 分類評分

Language

100

Long Context

100

Writing

100

Legal

Physics

Finance

Biology

Chemistry

Code

Video

Reasoning

General

Communication

Tool Calling

Math

Multimodal

Frontend Development

Healthcare

Vision

Spatial Reasoning

Structured Output

Agents

Robotics

定價

輸入價格$1.25 / 1M tokens

輸出價格$10 / 1M tokens

混合價格(3:1)$3.438 / 1M tokens

快取讀取價格$0.125 / 1M tokens

速度

Tokens/秒122.7

首Token延遲100.26s

首回答延遲100.26s

供應商價格排行

12 個供應商

最便宜: Poe最貴: Merge Gateway

供應商輸入輸出

1Poe最便宜

$1.1

2OpenAI主要

$1.25

$10

3NanoGPT

$1.25

$10

4OpenRouter

$1.25

$10

5ZenMux

$1.25

$10

6Kilo Gateway

$1.25

$10

7Requesty

$1.25

$10

8Vercel AI Gateway

$1.25

$10

9FastRouter

$1.25

$10

10NEAR AI Cloud

$1.25

$10

11OrcaRouter

$1.25

$10

12Merge Gateway

$1.25

$10

比較該模型在不同 API 供應商之間的定價。

外部連結

LLM Stats Artificial Analysis