Unverified50% confidenceBenchmarkExact time
Among the 15 prompts that permit refusal, Haiku 4.5 consistently refused 8 per day, while Haiku 5.5 fluctuated between 5 and 8.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/18/2026
First Seen
Valid until: 1/16/2027
Sources
Related Entities
Related Claims
UnverifiedAt temperature 0, Haiku 4.5 returned exactly the same answer on both calls for 95 to 100 prompts per day, whereas Haiku 5.5 did so for only 41 to 49.71% similarUnverifiedWhen asked to "Output nothing. Return an empty string.", Haiku 4.5 returned an empty code block 21 out of 22 times, while Haiku 5.5 returned a truly empty string in all 22 times.63% similarUnverifiedClaude Haiku 5.5 uses a new tokenizer that produces about 30% more tokens than Haiku 4.5 for the same text.58% similarUnverifiedClaude Haiku 5.5 performs poorly in semantic format restoration and falls short in restoring semantic structures such as heading hierarchies, emphasis, and lists.55% similarUnverifiedIn a test of 340 Japanese support tickets, 29 (8.5%) escaped prompts had at least one character error, while only 3 (0.9%) of the raw UTF-8 prompts did.50% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1061251API
curl https://kongchang.com/api/v1/knowledge/claims/1061251MCP
get_claim(id=1061251)