GPT-5.2 (Xhigh)
OpenAIGPTProprietary
描述
GPT‑5.2 introduces substantial gains in professional knowledge work, outperforming experts on GDPval with 70.9% wins or ties, and setting new highs in coding (SWE‑Bench Pro 55.6%), science (GPQA Diamond ~92–93%), math (AIME 2025: 100%), long‑context accuracy up to 256k tokens, and reliable tool‑calling (Tau2 Telecom 98.7%). It rolls out as Instant, Thinking, and Pro—faster, more structured, and less error‑prone—priced at $1.75/1M input and $14/1M output tokens, with Pro variants supporting xhigh reasoning for top‑quality, end‑to‑end execution.
发布日期
2025-12-11
参数规模
—
上下文长度
400K
支持模态
image, text
能力雷达图
80
general
70
coding
70
reasoning
77
science估算
70
agents
85
multimodal
缺少专门科学评测时,Science 由 LLM Stats 科学得分或推理能力估算。
排行榜排名
基准测试分数 (LLM Stats)
(LLM Stats (zeroeval))Agents
Toolathlon
46.3%自报
Communication
Tau2 Telecom
98.7%自报
Tau2 Retail
82.0%自报
Language
MMMLU
89.6%自报
Math
AIME 2025
100.0%自报
HMMT 2025
99.4%自报
LiveBench
74.8%
FrontierMath
40.3%自报
Multimodal
VideoMMMU
85.9%自报
Reasoning
Graphwalks BFS <128k
94.0%自报
GPQANYU + Cohere + Anthropic (2023)
92.4%自报
BrowseComp Long Context 128k
92.0%自报
BrowseComp Long Context 256k
89.8%自报
Graphwalks parents <128k
89.0%自报
ARC-AGI
86.2%自报
CharXiv-R
82.1%自报
SWE-Bench Verified
80.0%自报
SWE-Lancer (IC-Diamond subset)
74.6%自报
BrowseCompOpenAI (2025)
65.8%自报
MCP Atlas
60.6%自报
ARC-AGI v2
52.9%自报
Humanity's Last Exam
34.5%自报
Vision
ScreenSpot Pro
86.3%自报
MMMU-Pro
79.5%自报
AA 评测指数
(Artificial Analysis)暂无 AA 评测数据
LLM Stats 分类评分
(LLM Stats (zeroeval))Language90
Physics90
Grounding90
Healthcare90
Biology90
Chemistry90
Communication90
Multimodal80
Reasoning80
Search80
Spatial Reasoning80
Frontend Development80
General80
Math70
Code70
Tool Calling70
Vision70
Agents60
定价
输入价格$1.75 / 1M tokens
输出价格$14 / 1M tokens
混合价格(3:1)$4.8125 / 1M tokens
缓存读取价格$0.175 / 1M tokens
速度
Tokens/秒72.1
首Token延迟105.44s
首回答延迟105.44s
供应商价格排行
供应商价格排行
2 个供应商
最便宜: OpenAI最贵: Neon
供应商输入输出
1OpenAI主要
$1.75
$14
2Neon
$1.75
$14
比较该模型在不同 API 供应商之间的定价。