Expired70% confidenceFactTime unknown
The tester used Claude Code to conduct agentic coding tests on Claude 4.5 Haiku, and it performed extremely poorly on tasks including Movie Tracker, Go Terminal Calculator, Godot game development, and Svelte/Nuxt/Tauri frameworks.
1
Sources
70%
Confidence
Medium-term (~90 days)
Relevance
6/1/2026
First Seen
Valid until: 8/30/2026(expired)
Sources
Related Entities
Related Claims
UnverifiedAfter modifying code, Claude Code proactively calls the Bash tool to run unit tests and Lint checks, and when tests fail, it can read the Stack Trace and re-enter the modification loop.67% similarUnverifiedTesters believe that a certain gap still exists between Kimi K2.6 and foreign SOTA models such as Claude 4 in terms of Web Coding capability.66% similarUnverifiedA developer conducted a side-by-side comparison test of three AI programming tools—Claude Code, Gemini, and Cursor—for native Android application development.66% similarUnverified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% similarUnverifiedQwenCoder is a model designed specifically for code generation by Alibaba, and official benchmark results show its performance is comparable to Claude Sonnet 4.5.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/24353API
curl https://kongchang.com/api/v1/knowledge/claims/24353MCP
get_claim(id=24353)