Planning new generation capacity for real systems is a complex stochastic multistage multiscale MIP optimization problem. For example, planning horizons may be 25 years, due to long construction and licensing times for larger plans. In turn, the stochastic simulation of system operation requires weekly stages in order to represent uncertainties on demand, inflow and others. Finally, the operation within each stage uses hourly resolution due to the variability of renewable generation, unit commitment and ramping of thermal plants. In addition, it is necessary to co-optimize hourly generation reserve requirements and yearly capacity investment decisions. In this paper, we present PSR's planning model OptGen, built in Julia with JuMP framework, and applied to several large-scale studies such as the Brazilian and Chilean system.