The maximum number of tokens a language model can consider at once, influencing its text comprehension and generation capabilities. Determined by architectural and computational limits, it affects how models handle long texts and complex relationships.