I examine the optimal inflation target in a dynamic stochastic New Keynesian model featuring an occasionally binding zero lower bound on nominal interest rate (ZLB). To this end, I first calibrate the shock needed to generate the risk of hitting the ZLB that matches the U. S. data, based on a fully nonlinear method. I then resolve the model with different inflation targets and find that the optimal target is 3.4%. In addition, the optimal inflation target is a nonlinear function of the risk of hitting the ZLB and inflation indexation. It is always greater than 2% if the risk is greater than 2.5% or if the inflation indexation is higher than 0.5. Finally, the linear-quadratic approach overestimates the true optimal inflation target. In particular, based on the benchmark calibration, it generates an optimal target of 5.5%, compared with 3.4% found by the fully nonlinear method.