T · Token · tool use

Token

The unit of text language models consume and produce. Roughly 4 characters in English.

In one sentence

A token is the discrete unit of text — typically a word fragment of about 3-4 characters in English — that language models consume on input and produce on output, and which API providers charge per.

When it matters

When estimating context budget and cost. A 1,200-page PDF is ~600k tokens; a typical chat turn is ~600.

A real-world example

helpcode KB sends only the top-12 retrieved chunks (~6k tokens) plus the user’s message — keeps every reply under 8k tokens of input.

Curated by helpcode research team · Last reviewed 2026-05-22