Quadrotors are extremely agile, so much in fact, thatclassic first-principle-models come to their limits. Aerodynamiceffects, while insignificant at low speeds, become the dominantmodel defect during high speeds or agile maneuvers. Accuratemodeling is needed to design robust high-performance controlsystems and enable flying close to the platform’s physical limits.We propose a hybrid approach fusing first principles andlearning to model quadrotors and their aerodynamic effects withunprecedented accuracy. First principles fail to capture suchaerodynamic effects, rendering traditional approaches inaccuratewhen used for simulation or controller tuning. Data-drivenapproaches try to capture aerodynamic effects with blackboxmodeling, such as neural networks; however, they struggle torobustly generalize to arbitrary flight conditions. Our hybridapproach unifiesand outperformsboth first-principles blade-element momentum theory and learned residual dynamics. It isevaluated in one of the world’s largest motion-capture systems,using autonomous-quadrotor-flight data at speeds up to 65 km/h.The resulting model captures the aerodynamic thrust, torques,and parasitic effects with astonishing accuracy, outperformingexisting models with 50% reduced prediction errors, and showsstrong generalization capabilities beyond the training set.