Abstract Radiative transfer models (RTM) are used to calculate spectral and broadband irradiance, given a set of input parameters that are representative of the atmospheric state. While many studies exist on their accuracy, there is still a research gap in the assessment of their uncertainty, due to the nonlinear and not differentiable nature of the Radiative Transfer Equation, which is the core of a RTM. This study evaluates the uncertainty of both spectral and broadband irradiance calculated with the radiative transfer model SDISORT implemented in the tool UVSPEC within the range 280–2500 nm. A set of input values representing the atmospheric state at Kanzelhohe Observatory (Austria) site at 10:00 on April 25th, 2013 is taken as reference and a Monte Carlo technique is used to propagate the uncertainty of input parameters to the model output. Both the effects of single input parameter uncertainty and of their combination are evaluated, as well as the influence of the deviation of input values from the reference set. Results show that ozone column is an important source of uncertainty in the UV-B region, while the uncertainties of Angstrom aerosol turbidity coefficient and extraterrestrial spectrum affect the whole spectral range. Considering a reasonable variability range for all involved input parameters, the overall uncertainty of broadband global horizontal irradiance is between 2.9% and 5.9%. These values are higher, but still comparable, to typical uncertainty values of outdoor-deployed spectroradiometers.