Abstract: Transient mixed flow has occurred in the long tunnels of the water conveyance due to the rapid transition between free-surface and pressurized conditions. However, the simulation of the transient mixed flow can often lead to numerical instability and low computational efficiency. In this study, a first-order accurate implicit finite volume method (IFVM) was proposed to overcome these limitations using a central flux scheme and a temporal linearization technique. The updated cell-averaged hydraulic variables were evaluated through fluxes at the future time level. Thereby, the Courant–Friedrichs–Lewy (CFL) constraint was removed to limit the time step size in the conventional explicit schemes. A Taylor-series expansion and time linearization were applied to both flux and source terms, in order to avoid the iterative solutions for the high computational efficiency. The Saint-Venant equations were adopted as the governing equations. The Preissmann slot model was used to represent the free-surface and pressurized flow. Three classical laboratory experiments were carried out to verify the IFVM. Various transient behaviors were also covered, including hydraulic jumps and water-hammer wave propagation. Some parameters were measured on the arrival times of the flow transitions, the magnitude of pressure variations, and the locations of flow regime interfaces. The prediction was then in agreement with the measurement after tests. There were nearly identical water levels under Courant numbers of 5 and 10. The stability and accuracy were maintained even with the large time steps. A long-distance tunnel example was then simulated to evaluate the performance of IFVM, compared with the implicit finite difference method (IFDM) and the explicit finite volume method (EFVM). Among them, the tunnel was 10 km in length with a diameter of 10 m and a bottom slope of 0.001. Both direct and gradual mixed-flow transitions were triggered by different downstream gate discharges during simulation. Seriously spurious oscillations and mass-balance errors were produced with the IFDM during direct transitions. While the EFVM was often required to fully meet the much smaller time steps for high accuracy. By contrast, the IFVM provided the stable and smooth predictions for both transition types, where the computed water-surface profiles were nearly overlapped by those of the EFVM reference solution. The maximum water-volume error no more than 2.21%. In terms of computational efficiency, all models were executed on the same workstation that was equipped with a 4.70 GHz processor and 64 GB of memory. The average simulation time of the IFVM was 7 s, which was approximately 7 and 19 times faster than the IFDM (49 s) and EFVM (135 s), respectively. This improved performance was attributed to the implicit treatment of the fluxes and the non-iterative solution structure. The IFVM effectively captured the transient hydraulic processes, such as the hydraulic jumps, pressurization, and depressurization waves, particularly without suffering from numerical instability or excessive computation cost. Overall, the IFVM provided a robust and efficient numerical tool to simulate the transient mixed flows in the long water conveyance tunnels. Mass conservation was maintained to accurately reproduce the flow regime transitions, and then eliminate the time-step restrictions associated with traditional explicit schemes. The finding can also offer a promising alternative for the real-time analysis and operational scheduling of the large-scale water conveyance with both free-surface and pressurized flows.