DataLearner 标志

Claude Sonnet 4.5vsClaude Sonnet 4

在 26 个共同 benchmark 中,Claude Sonnet 4.5 整体领先:Claude Sonnet 4.5 领先 23 项,Claude Sonnet 4 领先 1 项,持平 2 项,平均分差 +14.79。

Anthropic
Claude Sonnet 4.5

Anthropic · 2025-09-30 · 聊天大模型

Anthropic
Claude Sonnet 4

Anthropic · 2025-05-23 · 推理大模型

Claude Sonnet 4.523 (88%)持平2(4%)1 Claude Sonnet 4

评测分数

按能力类目分组,每组内按分差大小排列;共 26 项。

General Knowledge

Claude Sonnet 4.5 领先 5/6
评测项Claude Sonnet 4.5Claude Sonnet 4分差
HLE33.6080 / 1729.60149 / 172+24
ARC-AGI63.7035 / 684049 / 68+23.70
ARC-AGI-213.6038 / 625.9046 / 62+7.70
MMLU Pro887 / 1328438 / 132+4
LiveBench53.6983 / 115Normal (No Tools)50.9889 / 115Normal (No Tools)+2.71
GPQA Diamond83.4063 / 18783.8062 / 187-0.40

Math and Reasoning

Claude Sonnet 4.5 领先 3/5
评测项Claude Sonnet 4.5Claude Sonnet 4分差
AIME20251001 / 1078551 / 107+15
FrontierMath - Tier 42.1056 / 80Normal (No Tools)072 / 80Normal (No Tools)+2.10
FrontierMath5.2038 / 604.1041 / 60+1.10
IMO-ProofBench27.108 / 1627.108 / 16持平
IMO-ProofBench Advanced4.806 / 84.806 / 8持平

Coding and Software Engineer

Claude Sonnet 4.5 领先 4/4
评测项Claude Sonnet 4.5Claude Sonnet 4分差
CodeClash1,3891 / 8Normal (With Tools)1,2234 / 8Normal (With Tools)+166
LiveCodeBench7148 / 1236659 / 123+5
SWE-bench Verified828 / 11280.2014 / 112+1.80
SWE-Bench Pro - Public43.6047 / 5442.7048 / 54+0.90

Agent Level Benchmark

Claude Sonnet 4.5 领先 2/2
评测项Claude Sonnet 4.5Claude Sonnet 4分差
τ²-Bench - Telecom985 / 356529 / 35+33
τ²-Bench84.709 / 435234 / 43+32.70

AI Agent - Tool Usage

Claude Sonnet 4.5 领先 2/2
评测项Claude Sonnet 4.5Claude Sonnet 4分差
OSWorld-Verified61.4020 / 2442.2022 / 24+19.20
Terminal-Bench503 / 3541.3010 / 35+8.70

Claw-style Agent Evaluation

Claude Sonnet 4.5 领先 2/2
评测项Claude Sonnet 4.5Claude Sonnet 4分差
Claw Bench88.1013 / 29Thinking (With Tools)77.8023 / 29Thinking (With Tools)+10.30
Pinch Bench88.204 / 37Thinking (With Tools)80.5022 / 37Thinking (With Tools)+7.70

Instruction Following

Claude Sonnet 4.5 领先 1/1
评测项Claude Sonnet 4.5Claude Sonnet 4分差
IF Bench57.3022 / 305523 / 30+2.30

Long Context

Claude Sonnet 4.5 领先 1/1
评测项Claude Sonnet 4.5Claude Sonnet 4分差
AA-LCR6610 / 156512 / 15+1

Multimodal Understanding

Claude Sonnet 4.5 领先 1/1
评测项Claude Sonnet 4.5Claude Sonnet 4分差
MMMU77.8015 / 2976.5017 / 29+1.30

Productivity Knowledge

Claude Sonnet 4.5 领先 1/1
评测项Claude Sonnet 4.5Claude Sonnet 4分差
GDPval-AA3916 / 213319 / 21+6

常识推理

Claude Sonnet 4.5 领先 1/1
评测项Claude Sonnet 4.5Claude Sonnet 4分差
Simple Bench54.3022 / 63Normal (No Tools)45.5034 / 63Thinking (No Tools)+8.80

规格对比

字段Claude Sonnet 4.5Claude Sonnet 4
发布机构AnthropicAnthropic
发布时间2025-09-302025-05-23
模型类型聊天大模型推理大模型
架构稠密模型稠密模型
参数规模暂无数据暂无数据
上下文长度1000K200K
最大输出64K64K

API 调用价格

价格优先使用 DataLearner 配置的 API 记录;缺失项不做推测。

价格项Claude Sonnet 4.5Claude Sonnet 4
文本输入$3 / 1M tokens$3 / 1M tokens
文本输出$15 / 1M tokens$15 / 1M tokens
缓存读取$0.3 / 1M tokens$0.3 / 1M tokens
缓存写入$3.75 / 1M tokens$3.75 / 1M tokens

小结

  • Claude Sonnet 4.5在以下类目领先:General Knowledge (5/6)、Math and Reasoning (3/5)、Coding and Software Engineer (4/4)、Agent Level Benchmark (2/2)、AI Agent - Tool Usage (2/2)、Claw-style Agent Evaluation (2/2)、Instruction Following (1/1)、Long Context (1/1)、Multimodal Understanding (1/1)、Productivity Knowledge (1/1)、常识推理 (1/1)

26 个共同 benchmark 上,Claude Sonnet 4.5 平均高出 14.79 分。

单项差距最大的 benchmark:CodeClash — Claude Sonnet 4.5 1,389,Claude Sonnet 4 1,223(分差 +166)。

本页正文由结构化模型、价格与 benchmark 数据生成,不使用实时 LLM 撰写。