A lightweight text-to-image model based on Qwen architecture, optimized for fast inference and efficient deployment.