Unverified50% confidenceFactExact time
Claude Haiku 5.5 uses a new tokenizer that produces about 30% more tokens than Haiku 4.5 for the same text.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/18/2026
First Seen
Valid until: 1/16/2027
Sources
Related Entities
Related Claims
UnverifiedClaude Haiku 5.5 performs poorly in semantic format restoration and falls short in restoring semantic structures such as heading hierarchies, emphasis, and lists.72% similarUnverifiedUnder Claude Haiku 5.5's tokenizer, per the official model overview, 1M tokens is roughly equal to 2.5M Unicode characters.66% similarUnverifiedWith the rule book, Haiku and Sonnet achieved identical scores, but Haiku is 14 times cheaper.62% similarUnverifiedThe Opus 4.7 tokenizer uses different merging strategies for code comments, mixed Chinese text, and special Unicode symbols, which may cause the token count for the same text to become approximately 1.0 to 1.7 times that of before.54% similarUnverifiedThe quick rule of thumb of dividing the English character count by 3 as a token estimate works roughly for English text, but significantly underestimates for CJK text such as Japanese, Chinese, and Korean, because the tokenizer splits CJK text at a finer granularity.54% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1060046API
curl https://kongchang.com/api/v1/knowledge/claims/1060046MCP
get_claim(id=1060046)