An information engine based on a two-level system in contact with a thermal reservoir is studied analytically. The model incorporates delay time between the measurement of the state of the system and the feedback. The engine efficiency and work extracted per cycle are studied as a function of delay time and energy spacing between the two levels. It is found that the range of delay time over which one can extract work from the information engine increases with temperature. For delay times comparable to the relaxation time, efficiency and work per cycle are maxima when k_{B}T≈2U_{0}, the energy difference between the levels. The generalized Jarzynski equality and the generalized integral fluctuation theorem are explicitly verified for the model. The results from the model are compared with the simulation results for a feedback engine based on a particle moving in a one-dimensional square potential. The variation of efficiency, work per cycle, and efficacy with the delay time is compared using relaxation time in the two-level model as the fitting parameter, leading to a good fit.