跳轉到主要內容

GPT-4o (Aug '24)

OpenAIGPTProprietary

描述

GPT-4o ('o' for 'omni') is a multimodal AI model that accepts text, audio, image, and video inputs, and generates text, audio, and image outputs. It matches GPT-4 Turbo performance on text and code, with improvements in non-English languages, vision, and audio understanding.

發布日期
2024-08-06
參數規模
上下文長度
128K
支援模態
image, pdf, text

能力雷達圖

8
general
32
coding
40
reasoning
36
science估算
50
agents
90
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287
31.0
AA
通用能力榜425
23.0
AA
數學推理196
46.0
AA
多模態榜30
81.0
LS
推理能力99
37.0
LS
科學能力315
36.0
AA

基準測試分數 (LLM Stats)

Biology

GPQA70.1%自報

Code

SWE-Bench Verified33.2%自報
SWE-Lancer32.6%自報
Aider-Polyglot30.7%自報
Aider-Polyglot Edit18.2%自報
SWE-Lancer (IC-Diamond subset)12.4%自報

Communication

Tau2 Retail63.4%自報
Multi-IF60.9%自報
TAU-bench Retail60.3%自報
Tau2 Airline45.5%自報
TAU-bench Airline42.8%自報
Multi-Challenge40.3%自報
Tau2 Telecom23.5%自報

Factuality

SimpleQA38.2%自報

Finance

MMLU85.7%自報
MMLU-Pro74.7%自報

General

MMMLU81.4%自報
IFEval81.0%自報
MMMU72.2%自報
MMMU-Pro59.9%自報
Internal API instruction following (hard)29.2%自報

Healthcare

VideoMMMU61.2%自報

Image To Text

DocVQA92.8%自報

Language

COLLIE61.0%自報

Long Context

EgoSchema72.2%自報
ComplexFuncBench66.5%自報
OpenAI-MRCR: 2 needle 128k31.9%自報

Math

MathVista61.4%自報
AIME 202413.1%自報
Humanity's Last Exam5.3%自報

Multimodal

AI2D94.2%自報
ChartQA85.7%自報
CharXiv-D85.3%自報
CharXiv-R58.8%自報

Reasoning

Graphwalks BFS <128k41.7%自報
Graphwalks parents <128k35.4%自報
ERQA35.2%自報

Video

ActivityNet61.9%自報

AA 評測指數

Intelligence Index
9.6
Math 500
0.8
Gpqa
0.5
Ifbench
0.4
Lcr
0.3
Scicode
0.3
Livecodebench
0.3
Tau2
0.3
Aime
0.1
Terminalbench Hard
0.1
Hle
0.0

LLM Stats 分類評分

Image To Text
90
Legal
80
Finance
80
Instruction Following
70
Language
70
Multimodal
70
Physics
70
Healthcare
70
Biology
70
Chemistry
70
Vision
70
Long Context
60
Structured Output
60
General
60
Writing
60
Math
50
Reasoning
50
Communication
50
Tool Calling
50
Spatial Reasoning
40
Factuality
40
Frontend Development
30
Code
30

定價

輸入價格$2.5 / 1M tokens
輸出價格$10 / 1M tokens
混合價格(3:1)$4.375 / 1M tokens
快取讀取價格$1.25 / 1M tokens

速度

Tokens/秒113.3
首Token延遲0.56s
首回答延遲0.56s

供應商價格排行

供應商價格排行

12 個供應商

最便宜: OpenAI最貴: Venice AI
供應商輸入輸出
1OpenAI最便宜
$0
$0.00001
2Azure
$0
$0.00001
3NanoGPT
$2.499
$9.996
4Abacus
$2.5
$10
5OpenRouter
$2.5
$10
6Kilo Gateway
$2.5
$10
7Cloudflare AI Gateway
$2.5
$10
8DigitalOcean
$2.5
$10
9Vercel AI Gateway
$2.5
$10
10OrcaRouter
$2.5
$10
11Merge Gateway
$2.5
$10
12Venice AI
$3.125
$12.5

比較該模型在不同 API 供應商之間的定價。

外部連結