The small time constants of power electronics devices lead to dynamic couplings with the electromagnetic transients of power networks, and thus complicate the modeling and stability analysis of power-electronics-based power systems. This paper presents a computationally-efficient approach to assess the small-signal stability of inverter-fed power systems. The power system is partitioned into individual components, including the power inverters, network impedances, and power loads. The state-space model of individual inverter is first built, where the frequency response and eigenvalue analysis collectively characterize the contributions of different controller parameters to the terminal behavior in a wide frequency range. These component models, together with the network equations, are then algebraically assembled based on the interconnection relations at their terminals. As a consequence, the state matrix of the whole system, which is essential to the system stability analysis, can be reformulated in a computationally-efficient way. The experimental results are finally given to validate the effectiveness of the modeling method and system stability analysis.