Expired85% confidenceEventTime unknown
A Bilibili UP host conducted systematic testing of Claude Haiku 4.5 across 5 visual programming tasks, with all results performing poorly
1
Sources
85%
Confidence
Medium-term (~90 days)
Relevance
6/3/2026
First Seen
Valid until: 9/1/2026(expired)
Sources
Claude Haiku 4.5 Real-World Test: Failed All 5 Programming Tasks
bilibilikarminski-牙医
Related Entities
Related Claims
UnverifiedThe tester rated Claude Haiku 4.5 as poor or very poor across five dimensions: 3D modeling comprehension, code performance optimization, library reference accuracy, physical logic reasoning, and instruction-following capability81% similarUnverifiedThe elephant toothpaste animation code generated by Claude Haiku 4.5 has serious performance bottlenecks and runs with significant lag68% similarUnverifiedThis evaluation was designed by Bilibili UP host '不正经的前端' and includes four coding scenarios of increasing difficulty.67% similarUnverifiedHaiku 4.5 has the lowest rate of misalignment behavior among all models in safety evaluations.65% similarUnverifiedIn testing, small models of approximately 1–2B parameters were unable to accurately locate and complete complex tool-calling tasks within lengthy system prompts.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/40309API
curl https://kongchang.com/api/v1/knowledge/claims/40309MCP
get_claim(id=40309)