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.

發布日期

2026-07-09

參數規模

—

上下文長度

1.1M

支援模態

image, pdf, text

能力雷達圖

general

coding

reasoning

science

agents

multimodal

排行榜排名

領域	#排名	分數	來源
智慧體能力模型榜	2	80.0	LS
程式碼能力榜	1	98.0	AA
通用能力榜	12	91.0	AA
多模態榜	10	66.0	LS
科學能力	5	94.0	AA

基準測試分數 (LLM Stats)

(LLM Stats (zeroeval))

Agents

Connectors

100.0%自報

Capture-the-Flag Challenges (Internal)

96.7%自報

Search and Function-Calling

91.0%自報

SEC-bench Pro

71.2%自報

Internal Research Debugging Evaluation

68.3%自報

KernelGen 1P

61.1%自報

LifeSciBench

59.9%自報

Toolathlon

58.0%自報

RSI Index

57.9%自報

Big Finance Bench

53.0%自報

Agents' Last Exam

52.7%自報

PostTrainBench Lite

50.3%自報

Management Consulting Tasks (Internal)

43.2%自報

ExploitGym

33.7%自報

GeneBench-Pro

28.7%自報

AutomationBench

18.1%自報

NanoGPT

9.7%自報

Code

BenchCAD (with Python tool)

83.4%自報

ExploitBench

73.5%自報

DeepSWE 1.1

73.0%

DeepSWE

72.7%自報

BenchCAD

70.6%自報

General

Artificial Analysis

59.0%

GDP.pdf

30.7%自報

Healthcare

HealthBench Consensus

95.5%自報

HealthBench Professional

60.5%自報

HealthBench

57.0%自報

HealthBench Hard

33.1%自報

Long Context

MRCR v2 (8-needle)

91.5%自報

MRCR v2 (8-needle, 512K-1M)

73.8%自報

Math

FrontierMath

89.0%自報

FrontierMath Tier 4 (v2)

83.0%自報

Multimodal

OSWorld 2.0

62.6%自報

Reasoning

GPQANYU + Cohere + Anthropic (2023)

94.6%自報

Graphwalks BFS >128k

90.7%自報

BrowseCompOpenAI (2025)

90.4%自報

Terminal-Bench 2.1

88.8%自報

Graphwalks BFS 1M

77.1%自報

SWE-Bench ProPrinceton NLP (2024)

64.6%自報

MedChemBench (Internal)

48.3%自報

FrontierCode 1.1

47.5%

ARC-AGI-3

7.8%自報

Vision

MMMU-Pro (with tools)

84.6%自報

MMMU-Pro

83.0%自報

AA 評測指數

(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 分類評分

(LLM Stats (zeroeval))

Math

Physics

Biology

Chemistry

Long Context

Spatial Reasoning

Multimodal

Safety

Tool Calling

Vision

Reasoning

General

Healthcare

Agents

Code

Science

Finance

Systems

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

快取寫入價格$6.25 / 1M tokens

速度

Tokens/秒78.3

首Token延遲76.98s

首回答延遲76.98s

供應商價格排行

2 個供應商

最便宜: OpenAI最貴: Neon

供應商輸入輸出

1OpenAI最便宜

$0.00001

$0.00003

2Neon

$30

比較該模型在不同 API 供應商之間的定價。

外部連結

LLM Stats Artificial Analysis