跳转到主要内容

GPT-4.5 (Preview)

OpenAIGPTProprietary

描述

GPT-4.5 is OpenAI's most advanced model, offering improved reasoning, coding, and creative capabilities with faster performance and longer context handling than GPT-4. It features enhanced instruction following, reduced hallucinations, and better factual accuracy.

发布日期
2025-02-27
参数规模
上下文长度
支持模态
image, text

能力雷达图

20
general
50
coding
80
reasoning
60
science估算
60
agents
70
multimodal

Science 在缺少专门科学评测时使用推理能力代理估算。

排行榜排名

领域#排名分数来源
通用能力榜378
28.0
AA
多模态榜47
75.0
LS
推理能力40
73.0
LS

基准测试分数 (LLM Stats)

Biology

GPQA69.5%自报

Code

HumanEval88.0%自报
Aider-Polyglot Edit44.9%自报
SWE-Bench Verified38.0%自报
SWE-Lancer37.3%自报
SWE-Lancer (IC-Diamond subset)17.4%自报

Communication

Multi-IF70.8%自报
TAU-bench Retail68.4%自报
TAU-bench Airline50.0%自报
Multi-Challenge43.8%自报

Factuality

SimpleQA62.5%自报

Finance

MMLU90.8%自报

General

IFEval88.2%自报
MMMLU85.1%自报
MMMU75.2%自报
Internal API instruction following (hard)54.0%自报

Language

COLLIE72.3%自报

Long Context

ComplexFuncBench63.0%自报
OpenAI-MRCR: 2 needle 128k38.5%自报

Math

GSM8k97.0%自报
MathVista72.3%自报
AIME 202436.7%自报

Multimodal

CharXiv-D90.0%自报
CharXiv-R55.4%自报

Reasoning

Graphwalks parents <128k72.6%自报
Graphwalks BFS <128k72.3%自报

AA 评测指数

Intelligence Index
20.0

LLM Stats 分类评分

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

定价

输入价格免费
输出价格免费
混合价格(3:1)免费

速度

Tokens/秒0.0 tokens/s
首Token延迟0.00s
首回答延迟0.00s

可用提供商

(LS 内部计价单位)

暂无提供商数据

外部链接