Mineral dust is a key player in the Earth system that affects the weather and climate through absorbing and scattering the radiation. Such effects strongly depend on the optical properties of the particles that are in turn affected by the particle shape. For simplicity, dust particles are usually assumed to be spherical. But this assumption can lead to large errors in modeling and remote sensing applications. This study investigates the impact of dust particle shape on its direct radiative effect in a next‐generation atmospheric modeling system ICON‐ART (ICOsahedral Nonhydrostatic weather and climate model with Aerosols and Reactive Trace gases) to verify if accounting for nonsphericity enhances the model‐observation agreement. Two sets of numerical experiments are conducted by changing the optical shape of the particles: one assuming spherical particles and the other one assuming a mixture of 35 randomly oriented triaxial ellipsoids. The simulations are compared to MISR (Multiangle Imaging Spectroradiometer), AERONET (Aerosol Robotic Network), and CALIPSO (Cloud‐Aerosol Lidar and Infrared Pathfinder Satellite Observation) observations (with focus on North Africa). The results show that consideration of particle nonsphericity increases the dust AOD (Aerosol Optical Depth) at 550 nm by up to 28% and leads to slight enhancement of the agreement between modeled and measured AOD. However, the model performance varies significantly when focusing on specific regions in North Africa. These differences stem from the uncertainties associated with particle size distribution and emission mechanisms in the model configuration. Regarding the attenuated backscatter, the simulated profile assuming nonsphericity differs by a factor of 2 to 5 from the experiment assuming spherical dust and is in a better agreement with the CALIPSO observations.