Tokens are chunks of text, not always full words, but pieces of them. For example, “cat” is one token, but “unbelievable” might be two. LLMs read and generate language one token at a time. More tokens = longer prompts, longer replies and a higher cost.
