INTRODUCTION: Newborn resuscitation is commonly performed in the presence of face mask leak. Leak is highly variable, pressure-dependent and often unrecognized. The effectiveness of resuscitation devices to deliver adequate inflations in the presence of leak is unknown. Bench models simulating continuous leak have the disadvantage of not accurately reflecting leak occurring during clinical resuscitation. A dynamic leak model based on pressure-release valves was thus developed. METHODS: This study investigates self-inflating bag (SIB) and T-piece resuscitator (TPR) ventilation performance in the presence of dynamic (DLM) compared to continuous (CLM) leak models in a bench study. Five predefined leak levels were tested for each leak model (0%-87%). Resuscitation devices were connected to a test lung (compliance 0.6 mL/cmH RESULTS: Three thousand six hundred inflations were analyzed. DLM showed a decrease in actual tidal volumes from 0%-87% leak with tidal volume differences (SIB 4.8 mL, TPR 2.9 mL), contrasting to minimal change for CLM (SIB -0.6 mL, TPR 0.3 mL). CLM demonstrated larger differences between patient interface and actual leak. The absolute difference at 60 inflations/min at 87% leak were SIB 37.5%, TPR 18.2% for CLM compared to SIB 4.6%, TPR 1.4% for DLM. CONCLUSIONS: CLM may underestimate the impact of resuscitation device performance with poor correlation between patient interface and actual delivered volume. DLM demonstrates several advantages with a more accurate representation of face mask leak and will prove useful in modeling all systems delivering PPV.