What counts as a token

A token in deep learning is basically a sentence broken off into mutliple segments for the AI model to understand. For example, inputting the string "Sunny days bring joy and warmth", into a tokenizer, it will break the sentence into an array of words and characters the AI understands (called tokens). So this would output an array like ["Sunn", "day", "bring", "joy", "and", "warmth"] where each item in this array counts as a token.