The number of text units (tokens) a model processes or generates; longer reasoning processes consume more tokens and may increase latency or cost.
Multi-step reasoning, logic puzzles, mathematical problem-solving