The Monte Carlo simulation landscape for the CMS experiment has been enriched in view of the challenges for the High Luminosity (HL)-LHC. In addition to the very different functions of Full Simulation and Fast Simulation, a new player, FlashSim, is gaining in importance. Full Simulation scores with a precise simulation based on the Geant4 detector simulation, but at the expense of runtime. The other tool, Fast Simulation, wins with significantly higher speed through the use of parametric particle-material interactions, but has the disadvantage of lower accuracy. New developments for these tools focus on compensating for these drawbacks to achieve faster or more accurate results. For the LHC a new decade of operation begins with a major upgrade, the HL-LHC. This requires also a significant upgrade of the CMS detector, whereby parts of the detector are replaced by new and more complex ones. The simulation toolkit Geant4 is constantly being improved to achieve faster and more accurate results, and the detector description must be expanded and adapted to the new parts and geometries. The new challenges are met with the use of machine learning (ML) techniques and processing jobs on graphics processing units (GPUs) to speed up the time-consuming simulations. FlashSim is an ML-based simulation framework that is trained for general use on a variety of different analyses and combines speed and accuracy. This contribution reports on the latest innovations and developments in the field of Full and Fast Simulation and points to other promising simulation tools such as FlashSim to fulfill the significantly increased requirements for the future of CMS.