Intranasal infection is commonly used to establish a SARS-CoV-2 mouse model due to its non-invasive procedures and a minimal effect from the operation itself. However, mice intranasally infected with SARS-CoV-2 have a high mortality rate, which limits the utility of this model for exploring therapeutic strategies and the sequelae of non-fatal COVID-19 cases. To resolve these limitations, an aerosolised viral administration method has been suggested. However, an in-depth pathological analysis comparing the two models is lacking. Here, we show that inhalation and intranasal SARS-CoV-2 (106 PFU) infection models established in K18-hACE2 mice develop unique pathological features in both the respiratory and central nervous systems, which could be directly attributed to the infection method. While the inhalation-infection model exhibited relatively milder pathological parameters, it closely mimicked the prevalent chest CT pattern observed in COVID-19 patients with focal, peripheral lesions and fibrotic scarring in the recuperating lung. We also found the evidence of direct neuron-invasion from the olfactory receptor neurons to the olfactory bulb in the intranasal model and showed the trigeminal nerve as an alternative route of transmission to the brain in inhalation infected mice. Even after viral clearance confirmed at 14 days post-infection, mild lesions were still found in the brain of inhalation-infected mice. These findings suggest that the inhalation-infection model has advantages over the intranasal-infection model in closely mimicking the pathological features of non-fatal symptoms of COVID-19, demonstrating its potential to study the sequelae and possible interventions for long COVID.