Model predictive control (MPC) is applied to the Caltech ducted fan, a thrust-vectored flight experiment. A real-time trajectory generation software based on spline theory and sequential quadratic programming is used to implement the MPC controllers. Timing issues related to the computation and implementation of repeatedly updated optimal trajectories are discussed. Results show computational speeds greater than 10 Hz, 2.5 times that of the actuator dynamics. The MPC controllers successfully stabilize a step disturbance applied to the ducted fan and compare favorably to LQR methods.