<p>In recent years, quadruped robots have made remarkable progress in dynamic mobility. However, most existing methods are still limited to specific motion patterns and lack a unified framework capable of flexibly adapting to predefined trajectories and gaits. To overcome these limitations, we propose a novel control framework that integrates nonlinear model predictive control (NMPC) with a hybrid whole-body control (WBC) strategy. NMPC is used to optimize complex gaits and highly nonlinear trajectories in real time through a switching cost and constraint mechanism. The hybrid WBC combines task prioritization with weight-based coordination, enabling concurrent execution of multiple motion tasks. We validate the proposed framework through simulations and real-world experiments on a quadruped robot. The system successfully executes a range of agile behaviors, including upright walking, handstands, and diagonal stepping. These behaviors are smooth and have a strong robustness against disturbances and delays. Our approach demonstrates, for the first time, that a unified NMPC-WBC controller can enable real-time execution of diverse and highly dynamic quadruped motions. The framework presents a significant step forward in achieving animal-level agility in legged robotic systems.</p>