A token is the atomic unit a language model reads and generates — typically a word, subword, or character, determined by the model's tokenizer. Most modern LLMs use subword tokenization (like BPE), which balances vocabulary size against the ability to represent rare words. Token count directly affects computational cost, context window limits, and pricing for API-based models. Understanding tokenization is essential for prompt engineering, cost estimation, and diagnosing unexpected model behavior with unusual inputs.