Unverified65% confidenceFactTime unknown

在实际性能测试中，GLM 4.6的表现远远优于Claude 4.5 Haiku

1

Sources

65%

Confidence

Medium-term (~90 days)

Relevance

6/1/2026

First Seen

Valid until: 8/30/2026

Sources

Claude 4.5 Haiku实测翻车：编码能力全面溃败，性价比被竞品碾压

bilibili烟神殿殿主

Related Entities

GLM 4.6model Claude 4.5 Haikumodel

Related Claims

UnverifiedGLM 5.2 在前端生成质量上优于 Claude Sonnet 572% similar Partially VerifiedClaude Haiku 4.5在SWE-bench基准测试中取得了73.3%的得分71% similar Unverified测评者认为GLM 5.2在纯编码上依然强劲，但涉及审美、物理感和模拟质感任务时HY3更胜一筹70% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/24356

API

curl https://kongchang.com/api/v1/knowledge/claims/24356

MCP

get_claim(id=24356)