When large language models generate code, output length and response time are affected by multiple factors such as model size, server load, prompt complexity, and context window limits.