Unverified50% confidenceFactExact time
OpenAI uses the BPE (Byte Pair Encoding) algorithm for tokenization, while Anthropic's Claude uses a similar but not identical tokenization strategy.
1
Sources
50%
Confidence
Long-term
Relevance
8/10/2026
First Seen
Sources
Rippling AI Spend Console：企业AI支出管理与ROI量化实战指南
Related Claims
UnverifiedThe Byte Pair Encoding (BPE) tokenization algorithm was originally introduced by OpenAI in GPT-2, splitting text into subword units by iteratively merging high-frequency character pairs.73% similarUnverified在OpenAI的tiktoken和Google的SentencePiece等现代分词器中，BPE已进化为BBPE（Byte-level BPE），直接在UTF-8字节层面操作70% similarUnverifiedIn English, approximately every 4 characters form 1 Token, and OpenAI uses the BPE (Byte Pair Encoding) algorithm to map high-frequency words to single Tokens.69% similarUnverifiedBPE（字节对编码）算法最初是一种数据压缩算法，后被引入NLP领域并应用于GPT、Claude等主流大语言模型的分词器设计69% similarVerifiedBPE（Byte Pair Encoding）是大语言模型使用的Token化算法，通过统计训练语料中字符对的共现频率迭代合并最常见的字符组合69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/725261API
curl https://kongchang.com/api/v1/knowledge/claims/725261MCP
get_claim(id=725261)