Submitted by X 139 Boogu-Image-0.1: Boosting Open Agentic Multimodal Generation via Understanding under a Minimal Budget Boogu 947 4