Tokenizer vocabulary

DeepSeek V3

DeepSeek V3 Hugging Face ByteLevel BPE tokenizer with 128k mergeable tokens, reserved control tokens, and non-special added-token metadata. Vocabulary size, token ranges, and special-token IDs are listed here.

DeepSeek

Creator

DeepSeek

Mergeable tokens

128,000

Total known tokens

128,804

Browse by type

PunctuationSymbols and marks WhitespaceSpaces, tabs, newlines EmojiPictographs and fragments MultilingualExamples across scripts CodeProgramming tokens Control/special tokensReserved tokenizer values

Open token index