Unverified70% confidenceFactTime unknown

测试者使用Claude Code对Claude 4.5 Haiku进行代理编码测试，在Movie Tracker、Go Terminal Calculator、Godot游戏开发、Svelte/Nuxt/Tauri框架等任务中均表现极差

1

Sources

70%

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

Claude 4.5 Haikumodel Claude Codeproduct Anthropiccompany

Related Claims

UnverifiedClaude Code改完代码后会主动调用Bash工具运行单元测试和Lint检查，测试失败时能读取Stack Trace并重新进入修改循环67% similar Unverified测试者认为Kimi K2.6与Claude 4等国外SOTA模型在Web Coding能力上仍存在一定差距66% similar Unverified一位开发者使用Claude Code、Gemini和Cursor三款AI编程工具进行Android原生应用开发的横向对比测试66% similar Unverified测试者采用 Codex 规划素材 + Claude Code 编码实现的分工模式开发类《杀戮尖塔》卡牌游戏65% similar UnverifiedQwenCoder是阿里专门为代码生成设计的模型，官方基准测试显示其性能可媲美Claude Sonnet 4.565% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/24353

API

curl https://kongchang.com/api/v1/knowledge/claims/24353

MCP

get_claim(id=24353)