The cost charged per token (unit of text) processed by a model, which varies based on model capability and complexity.