Kernel set in maize (Zea mays L.) hybrid seed production can be simulated using simple flowering characteristics of the male and female inbreds when pollen production is limiting. This approach, however, could overestimate kernel set when early kernel growth is limited by assimilate supply. We tested whether the accuracy of kernel set simulations could be improved using a plant growth model that incorporates both sink and source limitations for kernel set. Three models were compared: CERES-Maize, our kernel set model based on flowering characteristics, and a modified version of CERES-Maize that incorporated our kernel set model as a selectable module. Kernel set was simulated in six seed corn production fields located in Iowa. The mean grain yield simulated by CERES-Maize was 10 % less than the mean measured grain yield and it failed to simulate differences among fields (r2 = 0.51, MSD 19.5). The flowering model over-estimated kernel set by 4% on average, simulating with accuracy kernel set across a wide range of seed yields (r2 = 0.76, MSD 6.3). The modified version of CERES-Maize simulated kernel set within 1% of the mean actual kernel number (r2 = 0.77, MSD 6.2). The modified version of CERES-Maize did not simulate the flowering process as accurately as the flowering model because of a limitation in CERES-Maize sensitivity for phenology simulation. Incorporating a kernel set module into CERES-Maize provides a framework for simulating kernel set under sink or source limiting conditions.