This paper proposes a new direct model reference adaptive control algorithm for continuous time nonminimum phase systems. By introducing an auxiliary signal and a partial state vector, a minimal error model is constructed in state space. Using the error model an estimation algorithm for the unknown controller and auxiliary parameters is obtained via Lyapunov's direct method. For the fast convergence, speed-gradient approach is introduced. The global stability of the overall system is established under the persistency of excitation condition.