跳转到主要内容

Claude Opus 4.6 (Non-reasoning, High)

AnthropicClaudeProprietary

描述

Claude Opus 4.6 is Anthropic's most intelligent model, improving on its predecessor's coding skills with more careful planning, longer agentic task sustenance, more reliable operation in larger codebases, and better code review and debugging skills. First Opus-class model with 1M token context window (beta), 128K output tokens, and adaptive thinking. Features effort controls (low/medium/high/max) and context compaction for long-running tasks. State-of-the-art on Terminal-Bench 2.0, Humanity's Last Exam, GDPval-AA, and BrowseComp. Pricing: $5/$25 per million tokens (input/output).

发布日期
2026-02-05
参数规模
—
上下文长度
1.0M
支持模态
image, pdf, text

能力雷达图

100
general
70
coding
80
reasoning
77
science估算
80
agents
80
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

领域#排名分数来源
智能体能力模型榜41
57.0
LS
多模态榜25
61.0
LS

基准测试分数 (LLM Stats)

(LLM Stats (zeroeval))

Agents

Finance Agent60.7%自报

Code

CyberGym73.8%自报
FrontierSWE56.0%
OpenRCA34.9%自报

Communication

Tau2 Telecom99.3%自报
Tau2 Retail91.9%自报

Language

MMMLU91.1%自报

Legal

Legal Agent Benchmark4.2%

Long Context

MRCR v2 (8-needle)76.0%自报

Math

AIME 202599.8%自报
LiveBench76.3%

Multimodal

OSWorld72.7%自报

Reasoning

Vending-Bench 2801759.0%自报
Graphwalks parents >128k95.4%自报
GPQANYU + Cohere + Anthropic (2023)91.3%自报
BrowseCompOpenAI (2025)84.0%自报
SWE-Bench Verified80.8%自报
SWE-bench Multilingual77.8%自报
CharXiv-R77.4%自报
ARC-AGI v268.8%自报
Terminal-Bench 2.0Stanford × Laude Institute (2026)65.4%自报
MCP Atlas62.7%自报
Graphwalks BFS >128k61.5%自报
Humanity's Last Exam53.1%自报

Safety

FigQA78.3%自报

Search

DeepSearchQA91.3%自报

Vision

MMMU-Pro77.3%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))
Agents
100
Reasoning
100
General
100
Communication
100
Language
90
Physics
90
Search
90
Biology
90
Chemistry
90
Long Context
80
Math
80
Multimodal
80
Safety
80
Spatial Reasoning
80
Frontend Development
80
Healthcare
80
Tool Calling
80
Code
70
Vision
70
Finance
60
Legal
0

定价

输入价格$5 / 1M tokens
输出价格$25 / 1M tokens
混合价格(3:1)$10 / 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38.8
首Token延迟1.72s
首回答延迟1.72s

供应商价格排行

供应商价格排行

1 个供应商

供应商输入输出
1Anthropic主要
$5
$25

比较该模型在不同 API 供应商之间的定价。

外部链接