Time-Series Language Models (TSLMs) are foundation models extended with a time-series modality, enabling direct reasoning, explanation, and forecasting over temporal data in natural language.