Expired65% confidenceFactExact time
Sonnet 5 uses a new tokenizer that consumes more tokens for the same content, with the official figure being 1.0 to 1.35 times that of Sonnet 4.6.
3
Sources
65%
Confidence
Medium-term (~90 days)
Relevance
7/22/2026
First Seen
Valid until: 10/20/2026(expired)
Sources
Related Claims
UnverifiedAnthropic 在 Opus 4.7 中引入了新分词器，Sonnet 5 沿用该分词器，同样文本约多 30% 的 Token72% similarUnverifiedSonnet 5.5 tends to use British spelling when writing (e.g., spelling "color" as "colour"), but this can be corrected with explicit instructions.68% similarUnverifiedThe ratio of bytes to tokens varies greatly across different tokenizers and character sets: English text is about 4 bytes/token, Chinese about 1.5–3 bytes/token, and JSON Schema commonly falls between 3.5–4.5.54% similarUnverifiedWhen T5 does summarization it typically uses abstractive generation, i.e. reorganizing the language rather than extracting original sentences, so it must work in tandem with transcription quality—typos and spoken-word redundancy in the transcription directly affect the coherence of the summary.50% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/593950API
curl https://kongchang.com/api/v1/knowledge/claims/593950MCP
get_claim(id=593950)