A developer is deploying a generative AI model to produce short product summaries for an e-commerce catalog. Each summary must stay concise, and the team wants to control costs by capping how much text the model generates per request. Which configuration parameter should the developer adjust to limit the length of each generated response?