This paper presents a direct identification algorithm for continuous-time low-order models both with delays and free of delays based on the method of weighted moments matching. In this approach, it is sufficient to obtain response data stimulated by a single pulse, such as a rectangular or a half-sine pulse. From the weighted time moments of both input and output data, the parameters of low-order transfer function models are determined by solving linear equations in a matrix form. Numerical examples show that the proposed algorithm provides robust identification results despite the fact that the system considered is oscillatory with a long tail and contains measurement noise.