跳轉到主要內容

GPT-5.5 (Xhigh)

OpenAIGPTProprietary

描述

GPT-5.5 is OpenAI's smartest model yet, designed for real work across agentic coding, computer use, knowledge work, and early scientific research. It matches GPT-5.4 per-token latency in real-world serving while reaching a much higher level of intelligence and using significantly fewer tokens to complete the same tasks. GPT-5.5 supports a 1M-token context window in the API and a 400K-token context window in Codex, with state-of-the-art results on Terminal-Bench 2.0, OSWorld-Verified, GDPval, FrontierMath, and CyberGym.

發布日期
2026-04-23
參數規模
—
上下文長度
1.1M
支援模態
image, pdf, text

能力雷達圖

70
general
70
coding
60
reasoning
77
science估算
80
agents
85
multimodal

缺少專門科學評測時,Science 由 LLM Stats 科學得分或推理能力估算。

排行榜排名

領域#排名分數來源
智慧體能力模型榜56
50.0
LS
程式碼能力榜23
85.0
LB
數學推理8
96.0
LB
多模態榜28
61.0
LS
推理能力15
90.0
LB

基準測試分數 (LLM Stats)

(LLM Stats (zeroeval))

Agents

BixBench80.5%自報
Finance Agent60.0%自報
Toolathlon55.6%自報
OfficeQA Pro54.1%自報
Finance Agent v251.8%
GeneBench25.0%自報

Code

CyberGym81.8%自報
FrontierSWE73.0%
DeepSWE 1.167.0%

Communication

Tau2 Telecom98.0%自報

General

GDPval-MM84.9%自報

Legal

Legal Agent Benchmark2.1%

Long Context

MRCR v2 (8-needle)74.0%自報

Math

LiveBench80.7%
FrontierMath35.4%自報

Multimodal

OSWorld-Verified78.7%自報

Reasoning

ARC-AGI95.0%自報
GPQANYU + Cohere + Anthropic (2023)93.6%自報
ARC-AGI v285.0%自報
BrowseCompOpenAI (2025)84.4%自報
Terminal-Bench 2.0Stanford × Laude Institute (2026)82.7%自報
MCP Atlas75.3%自報
SWE-Bench ProPrinceton NLP (2024)58.6%自報
Graphwalks parents >128k58.5%自報
Humanity's Last Exam52.2%自報
Graphwalks BFS >128k45.4%自報
FrontierCode 1.143.0%

Vision

MMMU-Pro83.2%自報

AA 評測指數

(Artificial Analysis)

暫無 AA 評測資料

LLM Stats 分類評分

(LLM Stats (zeroeval))
Communication
100
Physics
90
Biology
90
Chemistry
90
Multimodal
80
Safety
80
Search
80
Tool Calling
80
Vision
80
Spatial Reasoning
70
Finance
70
General
70
Code
70
Long Context
60
Math
60
Reasoning
60
Agents
60
Science
20
Legal
0

定價

輸入價格$5 / 1M tokens
輸出價格$30 / 1M tokens
混合價格(3:1)$11.25 / 1M tokens
快取讀取價格$0.5 / 1M tokens

速度

Tokens/秒77.5
首Token延遲38.88s
首回答延遲38.88s

供應商價格排行

供應商價格排行

4 個供應商

最便宜: Kunavo最貴: Neon
供應商輸入輸出
1Kunavo最便宜
$2
$12
2OpenAI主要
$5
$30
3FrogBot
$5
$30
4Neon
$5
$30

比較該模型在不同 API 供應商之間的定價。

外部連結