As evidenced by the Chelyabinsk and Tunguska airburst events in Russia, decameter-scale Near-Earth Objects (NEOs) can pose a hazard to human life and infrastructure from the energy they deposit in the atmosphere as they break up. To understand the potential damage these small NEOs can cause on Earth's surface, it is imperative to be able to model their atmospheric entry quickly and accurately. Here we compare three semi-analytical models of asteroid airbursts that differ in their descriptions of fragment separation and spreading. Each model can be calibrated to produce a good fit to the energy deposition curve inferred from Chelyabinsk observations, but in each case the implied initial meteoroid strength is different and when the calibrated models are upscaled to Tunguska, the results diverge. This introduces an inter-model uncertainty that compounds the large range of uncertain physical and model parameters that influence probabilistic hazard assessment. Uncertainty quantification of airburst energy deposition was performed for a theoretical impacting object with H-magnitude 27, assuming no prior knowledge of any other impactor or model parameter. Each of the three models produces a different distribution of airburst outcomes, however, the variation attributable to physical parameter uncertainty is far larger than the inter-model differences. To constrain the initial conditions of the Tunguska event, the same uncertainty quantification was performed for an H-magnitude 24 event. Among the scenarios consistent with Tunguska observations (5–10 km burst altitude, 10–60° trajectory angle, 3–50 MT TNT total energy release) the most likely range of impact conditions was: radius of 25–75 m, mass of 1× 10 8 – 2.5× 10 9 kg, initial velocity of 11.5–33 km/s, and angle of 25–60°.