A comprehensive model comparison test was conducted to evaluate the performance of six existing river ice hydraulic numerical models: CRISSP2D, ICESIM, HEC RAS, River1D, RIVICE, and VARY ICE. These 2D and 1D models were set up to simulate ice jam events on a 73.5 km stretch of the Athabasca River at Fort McMurray, Alberta. Three scenarios were simulated and outputs from each model were compared for i) the 1986 ice jam, ii) the 2020 ice jam and iii) a 1:100 AEP ice jam backwater profile. All modelling teams used the same geometric and hydrometric datasets and independently calibrated their models for each scenario. All six models successfully reproduced the observed backwater profile for both the 1986 and 2020 ice jams. The greatest differences were found near the jam toe where complex hydraulics due to islands and overbank spill led to more noticeable variations between the models. For the third scenario, the design event, each model was “calibrated” such that its simulated backwater profile matched the independently derived 1:100 AEP water level elevation at the Clearwater tributary confluence, using the 2020 calibration parameter set and suitably selecting the value of the unknown Athabasca River discharge. Although agreement at the location was consistently achieved by all models, the simulated profiles diverged progressively upstream and downstream of this point. Variations in the profiles have important implications for flood hazard assessment, as they may translate into large differences in floodplain depths, inundation extents, and dyke-crest elevation requirements along a river channel.