For text models, a tokenizer converts characters into a sequence of tokens with corresponding numeric identifiers. The same visible text can produce different token sequences under different tokenizers, so a token count is not a universal word count.
Token usage helps describe context size, processing and some pricing arrangements. Total task cost also depends on which tokens are charged, repeated requests and whether the result meets the goal. Counting fewer tokens does not by itself establish a better or cheaper completed outcome.
ELI5
Think of tokens as the pieces a model uses to read and write. A long word may be several pieces, while a common short word may be one.
For example, two tools can divide the same sentence into different pieces, so their token counters can disagree even though the words look identical.
Is one token always one word?
No. A token can represent a word, part of a word, punctuation or another encoded unit, depending on the tokenizer.
Does a lower token count mean a lower task cost?
Not necessarily. Prices, repeated attempts and the quality of the accepted result all affect the cost of completing a task.

