Using the North American Multi-Model Ensemble (NMME) hindcasts and ERSST observations, we investigate how the El Nino-Southern Oscillation (ENSO) prediction skill changes with the NMME model update and its possible causes. The NMME is divided into four phases of model update, approximately representing the operational model suites of 2011-2012, 2012-2014, 2014–2019, and 2019–present. ENSO prediction skills are analyzed in the hindcast for the period of 1982-2010. Results show that ENSO prediction skill does not change much in short-term prediction but improves significantly in long-term prediction with the model update. The possible reasons for the improvement are the improvement of ENSO Spring Persistence Barrier and the improvement of NMME's prediction skill for Atlantic Nino/Nina.