The world is a very big place, and 'Hilly' is not a very precise geomorphological term. If you gave the bounding box of your area of interest one might be able to give some advice.

And my advice to you is, to think about your objective and the objects or phenomenon you want to look it.

Afterwards you can look at different sources and their spatial resolution with which you can observe in your research area.

By their very nature, hilly areas are characterized by their high undulations. Therefore, the two deciding factors for 3D topographic modeling are the density of points or the spatial resolution of the DEM and the quality of height measurement. The best approach is to conduct a controlled experiment by having a high quality reference data and other sources of DEMs as the ones stated in your question and then compare the root mean square errors from different sources. The reference data can be obtained, for example, from drone mapping using Photogrammetric techniques.

The other respondents make excellent points - what sort of model and what is the goal? Inventory, hazard, prediction, monitoring, etc. may require a variety of inputs, and depending, whatever coarse DEM available may or may not a primary feed for the model. Also, the two choices you mentioned are certainly not the only possible sources of data, there are many other specialized elevation surveys for specific regions of interest.

In addition to the responses from fellow researchers, I would recommend the ALOS Global Digital Surface Model "ALOS World 3D (ALOS DSM) and the TanDEM-X data.

The ALOS World 3D - 30m (AW3D30) is a global digital surface model (DSM) dataset with a horizontal resolution of approximately 30 meters (1 arcsec mesh). The dataset is based on the DSM dataset (5-meter mesh version) of the World 3D Topographic Data.