Model risk has an important effect on risk measurements. Indeed, the choice of the underlying probabilistic model can have a significant impact on the risk forecast. The hazard of producing poor risk assessments due to the choice of an unsuited model is known as “model risk”. Its detection and quantification are crucial tasks, particularly with energy commodities which require more complex modelling compared to the ones needed in traditional financial markets. Using a normalized measure of model risk for the forecast of daily Value-at-Risk, we focus on a restricted set of plausible models within the GARCH-type class specified with nine different distributions. In this way, we are able to provide a more reliable assessment of model risk for two energy commodities (natural gas and crude oil) over the years from 2001 to 2015.