跳转到主要内容
AITier

GPT-5.6 Terra (max)

OpenAIGPTProprietary

描述

GPT-5.6 Terra is the balanced tier of OpenAI's GPT-5.6 family, designed for workloads that balance intelligence and cost. It delivers performance competitive with GPT-5.5 at roughly half the price, supports max reasoning effort, and has a 1.05M-token context window.

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

54
general
73
coding
93
reasoning
69
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13
63.0
LS
代码能力榜10
95.0
AA
通用能力榜18
88.0
AA
多模态榜38
57.0
LS
科学能力21
88.0
AA

基准测试分数 (LLM Stats)

(LLM Stats (zeroeval))

Agents

Connectors100.0%自报
Search and Function-Calling94.6%自报
Capture-the-Flag Challenges (Internal)91.8%自报
Internal Research Debugging Evaluation67.8%自报
SEC-bench Pro57.7%自报
RSI Index56.3%自报
LifeSciBench56.0%自报
Toolathlon53.1%自报
PostTrainBench Lite51.5%自报
Big Finance Bench51.0%自报
Agents' Last Exam50.4%自报
KernelGen 1P49.2%自报
Management Consulting Tasks (Internal)37.2%自报
GeneBench-Pro23.3%自报
ExploitGym23.2%自报
AutomationBench15.2%自报
NanoGPT14.5%自报

Code

BenchCAD (with Python tool)78.2%自报
DeepSWE 1.170.0%
DeepSWE69.6%自报
BenchCAD62.3%自报
ExploitBench52.9%自报

General

Artificial Analysis55.0%
GDP.pdf24.7%自报

Healthcare

HealthBench Consensus95.1%自报
HealthBench Professional57.7%自报
HealthBench57.0%自报
HealthBench Hard32.7%自报

Long Context

MRCR v2 (8-needle)89.6%自报
MRCR v2 (8-needle, 512K-1M)72.5%自报

Math

FrontierMath84.9%自报
FrontierMath Tier 4 (v2)68.3%自报

Multimodal

OSWorld 2.050.2%自报

Reasoning

GPQANYU + Cohere + Anthropic (2023)92.9%自报
BrowseCompOpenAI (2025)87.5%自报
Terminal-Bench 2.187.4%自报
Graphwalks BFS >128k76.9%自报
Graphwalks BFS 1M71.2%自报
SWE-Bench ProPrinceton NLP (2024)63.4%自报
FrontierCode 1.141.3%
MedChemBench (Internal)35.0%自报
ARC-AGI-30.8%自报

Vision

MMMU-Pro (with tools)82.0%自报
MMMU-Pro80.7%自报

AA 评测指数

(Artificial Analysis)
Coding Index(Artificial Analysis)
76.7
Intelligence Index(Artificial Analysis)
56.6
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))
0.6
Scicode(UIUC + Argonne National Lab (2024))
0.5
Hle(Center for AI Safety + Scale AI (2025))
0.4
Tau Banking
0.4

LLM Stats 分类评分

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

定价

输入价格$2 / 1M tokens
输出价格$12 / 1M tokens
混合价格(3:1)$4.5 / 1M tokens
缓存读取价格$0.2 / 1M tokens
缓存写入价格$2.5 / 1M tokens

速度

Tokens/秒119.1
首Token延迟88.13s
首回答延迟88.13s

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
$0.00001
2Neon
$2.5
$15

比较该模型在不同 API 供应商之间的定价。

外部链接

LLM StatsArtificial Analysis