Unverified85% confidenceFactTime unknown

Claude Haiku 4.5的错误拒绝率低于Sonnet 4.5和Opus 4.1，是Anthropic迄今最安全的模型之一

1

Sources

85%

Confidence

Medium-term (~90 days)

Relevance

6/1/2026

First Seen

Valid until: 8/30/2026

Sources

Claude Haiku 4.5深度解析：速度翻倍成本降三分之二的性价比之王

bilibili秒如科技

Related Entities

Claude Haiku 4.5model Claude Sonnet 4.5model Opus 4.1model Anthropiccompany

Related Claims

UnverifiedClaude Haiku 4.5在部分计算机使用任务基准测试中超越了更大型的Sonnet 4.576% similar Unverified根据Anthropic发布的基准测试数据，Sonnet 4.5在编码能力上略优于Claude Opus 4.175% similar UnverifiedAnthropic表示Sonnet 4.6在提示注入抵抗能力方面的表现与Opus 4.6相当74% similar Unverified在天气网站动效测试中，Claude Haiku 4.5的CSS动效细节表现超越了Claude Sonnet 4.072% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/24451

API

curl https://kongchang.com/api/v1/knowledge/claims/24451

MCP

get_claim(id=24451)