Base32 is binary-to-text encoding based on the base-32 numeral system. It uses an alphabet of 32 digits, each of which represents a different combination of 5 bits (25). Since base32 is not very widely adopted, the question of notation i.e. which characters to use to represent the 32 digits is not as settled as in the case of more well-known numeral systems (such as hexadecimal) even though RFCs and unofficial and de facto standards exist. One way to represent Base32 numbers in human-readable form is using digits 0–9 followed by the twenty-two upper-case letters A–V. However, many other variat
Base32 is binary-to-text encoding based on the base-32 numeral system. It uses an alphabet of 32 digits, each of which represents a different combination of 5 bits (25). Since base32 is not very widely adopted, the question of notation i.e. which characters to use to represent the 32 digits is not as settled as in the case of more well-known numeral systems (such as hexadecimal) even though RFCs and unofficial and de facto standards exist. One way to represent Base32 numbers in human-readable form is using digits 0–9 followed by the twenty-two upper-case letters A–V. However, many other variations are used in different contexts. Historically, Baudot code could be considered a modified (stateful) base32 code. Base32 is often used to represent byte strings.
== RFC 4648 encodings == The October 2006 proposed Internet standard documents base16, base32 and base64 encodings. It includes two schemes for base32, but recommends one over the other. It further recommends that regardless of precedent, only the alphabet it defines in its section 6 actually be called base32, and that the other similar alphabet in its section 7 instead be called base32hex. Agreement with those recommendations is not universal. Care needs to be taken when using systems that are called base32, as those systems could be base32 per RFC 4648 §6, or per §7 (possibly disregarding that RFC's deprecation of the simpler name for the latter), or they could be yet another encoding variant, see further below.
Discovered by embedding cosine similarity (sentence-transformers MiniLM, 384-dim).