跳转到主要内容

GPT-5.6 Sol (max)

OpenAIGPTProprietary

描述

GPT-5.6 Sol is the frontier model in OpenAI's GPT-5.6 family, designed for complex professional work across coding, knowledge work, cybersecurity, and science. It sets state-of-the-art results while using fewer tokens at lower estimated cost, supports max reasoning effort and an ultra multi-agent mode, and has a 1.05M-token context window. The gpt-5.6 alias routes to GPT-5.6 Sol.

发布日期
2026-07-09
参数规模
上下文长度
1.1M
支持模态
image, pdf, text

能力雷达图

58
general
74
coding
94
reasoning
72
science
70
agents
85
multimodal

排行榜排名

领域#排名分数来源
智能体能力模型榜2
80.0
LS
代码能力榜1
98.0
AA
通用能力榜12
91.0
AA
多模态榜10
66.0
LS
科学能力5
94.0
AA

基准测试分数 (LLM Stats)

(LLM Stats (zeroeval))

Agents

Connectors100.0%自报
Capture-the-Flag Challenges (Internal)96.7%自报
Search and Function-Calling91.0%自报
SEC-bench Pro71.2%自报
Internal Research Debugging Evaluation68.3%自报
KernelGen 1P61.1%自报
LifeSciBench59.9%自报
Toolathlon58.0%自报
RSI Index57.9%自报
Big Finance Bench53.0%自报
Agents' Last Exam52.7%自报
PostTrainBench Lite50.3%自报
Management Consulting Tasks (Internal)43.2%自报
ExploitGym33.7%自报
GeneBench-Pro28.7%自报
AutomationBench18.1%自报
NanoGPT9.7%自报

Code

BenchCAD (with Python tool)83.4%自报
ExploitBench73.5%自报
DeepSWE 1.173.0%
DeepSWE72.7%自报
BenchCAD70.6%自报

General

Artificial Analysis59.0%
GDP.pdf30.7%自报

Healthcare

HealthBench Consensus95.5%自报
HealthBench Professional60.5%自报
HealthBench57.0%自报
HealthBench Hard33.1%自报

Long Context

MRCR v2 (8-needle)91.5%自报
MRCR v2 (8-needle, 512K-1M)73.8%自报

Math

FrontierMath89.0%自报
FrontierMath Tier 4 (v2)83.0%自报

Multimodal

OSWorld 2.062.6%自报

Reasoning

GPQANYU + Cohere + Anthropic (2023)94.6%自报
Graphwalks BFS >128k90.7%自报
BrowseCompOpenAI (2025)90.4%自报
Terminal-Bench 2.188.8%自报
Graphwalks BFS 1M77.1%自报
SWE-Bench ProPrinceton NLP (2024)64.6%自报
MedChemBench (Internal)48.3%自报
FrontierCode 1.147.5%
ARC-AGI-37.8%自报

Vision

MMMU-Pro (with tools)84.6%自报
MMMU-Pro83.0%自报

AA 评测指数

(Artificial Analysis)
Coding Index(Artificial Analysis)
77.4
Intelligence Index(Artificial Analysis)
60.9
Gpqa(NYU + Cohere + Anthropic (2023))
0.9
Terminalbench V2 1
0.9
Tau2(Sierra + U Toronto + Vector Institute (2025))
0.9
Lcr(Artificial Analysis)
0.8
Ifbench(Google Research (2023))
0.7
Terminalbench Hard(Stanford × Laude Institute (2026))
0.7
Scicode(UIUC + Argonne National Lab (2024))
0.6
Hle(Center for AI Safety + Scale AI (2025))
0.5
Tau Banking
0.4

LLM Stats 分类评分

(LLM Stats (zeroeval))
Math
90
Physics
90
Search
90
Biology
90
Chemistry
90
Long Context
80
Spatial Reasoning
80
Multimodal
70
Safety
70
Tool Calling
70
Vision
70
Reasoning
60
General
60
Healthcare
60
Agents
60
Code
60
Science
50
Finance
50
Systems
40

定价

输入价格$5 / 1M tokens
输出价格$30 / 1M tokens
混合价格(3:1)$11.25 / 1M tokens
缓存读取价格$0.5 / 1M tokens
缓存写入价格$6.25 / 1M tokens

速度

Tokens/秒78.3
首Token延迟76.98s
首回答延迟76.98s

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最便宜
$0.00001
$0.00003
2Neon
$5
$30

比较该模型在不同 API 供应商之间的定价。

外部链接