Unverified50% confidenceBenchmarkExact time
According to a ComputingForGeeks evaluation, Claude Haiku 5.5 over-refuses more than other models in automated behavior audits, and in the Gray Swan coding scenario its prompt injection success rate dropped from Haiku 4.5's 58.4% to 0.08%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/19/2026
First Seen
Valid until: 1/17/2027
Sources
Related Entities
Related Claims
UnverifiedClaude Haiku 4.5 had a 20.2% capitulation rate under the first incorrect pressure but only a 7.6% rate of being broken by repetition; Claude Sonnet 5 resisted 94.4% of the first pressure attempts but lost 15.7% of those under repetition.69% similarUnverified4 basic rules can reduce the AI programming error rate from 41% to 11%66% similarUnverifiedUsing a Haiku-level model for simple classification tasks can reduce the per-call cost to as low as 1/20 of that of a flagship model or even lower.65% similarUnverifiedIn the 5-agent conversational cluster architecture (unstructured Markdown handoffs), the spec drift failure rate is 48.4%, the silent tool error propagation rate is 79.0%, the premature termination rate is 34.2%, and the end-to-end success rate is 18.5%.64% similarUnverifiedUnder the 1024-token limit, 49 of Haiku 5.5's responses (2.2%) were truncated, which never happened with Haiku 4.5; 6 of these came from the prompt "Explain the word 'time' in exactly 50 words".64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1067654API
curl https://kongchang.com/api/v1/knowledge/claims/1067654MCP
get_claim(id=1067654)