Hot forging of Inconel 718 is an essential process to give shape and in-use properties to final part design. Dynamic recrystallization is the main mechanism influencing the properties through a grain size control. It is then of crucial importance to be able to predict the mechanical behavior during forging and the microstructure evolution during and after the dynamic recrystallization. A coupling between a crystal plasticity finite element model and a recrystallization model using a 3D cellular automata approach is applied to a 3D polycrystalline aggregate deformed by compression. Beyond the prediction of the stress–strain curves at different strain rates, numerical simulations enable the visualization of stress, strain, energy, local orientation, dislocation density fields within the aggregate. Finally, the recrystallization kinetic and the grain size are recorded as a function of strain. The results are in a rather good agreement with experiments, even if the predicted grain size is slightly larger than the measured one. To improve the prediction, annealing twinning is accounted for.