The unit models actually read and bill for, and why character counts mislead you.
3 articles
Listed inTokensHow LLMs Workon
Why character and word counts are wrong by enough to matter, how BPE actually splits your text, and how to budget a request so the model never gets cut off mid-sentence.
Listed inTokensHow LLMs Workon
The BPE paper — why tokenizers merge frequent byte pairs, and why your rare identifiers cost so much.
Listed inTokensHow LLMs Workon
The unit models actually read and bill for, and why character counts mislead you.