Many endemic, emerging, re-emerging or potentially emerging RNA viruses (Flavi-, Alpha-, Henipa-, Rhabdo-, Corona-viruses…) target the human central nervous system (CNS), causing severe neurological disorders, sometimes fatal or leading to debilitating consequences. The Flaviviruses genus includes many of these highly pathogenic neurotropic viruses (West Nile virus, Tick-Borne Encephalitis virus-TBEV, Zika virus…). Despite their dangerousness, only a few vaccines exist and there is currently no available antiviral treatment. Whereas many efforts are made for the identification of antiviral molecules, including some with “broad-spectrum” properties, most studies use models that are not physiologically relevant. This probably gives an explanation to the lack of antiviral activity or excessive toxicity often observed in clinical trials, leading to their failure. To overcome this problem, we believe it is important to use pathological models of infection to identify molecules with high predictive value of therapeutic efficiency in vivo. Here, we used an in vitro two-dimensional (2D) culture of neuronal/glial cells differentiated from fetal human neural progenitor cells which reproduces major hallmarks of TBEV infection in the human brain. We showed that some molecules selected from the literature for their antiviral activity against TBEV or other Flaviviruses when using cell lines (VERO, A549, Huh-7 cells…) are not efficient when testing them in TBEV-infected neuronal/glial cells (hNGCs). On the contrary, we identified several molecules with an anti-TBEV activity in hNGCs that had no antiviral activity in a TBEV-infected cell line. Finally, we confirmed the antiviral activity of some of the molecules previously identified in cell lines. These results thus clearly show that antiviral activity depends on the cellular models used. They call for developing more physiologically relevant 2D models for testing or screening drugs for their antiviral activity. In an attempt to further improve our in vitro models of TBEV infection, we are currently using 3D-cerebral organoids and will soon test selected drugs for their efficiency and toxicity in this model. We hope that this will allow us to select compounds with a very high probability to be efficient and non-toxic in the human brain.