We end up with the following results for the two models:
l1 | 26570.0 | 28880.4 | 27287.8 | 29815.7 | 27822.0]L2 | 36670.9 | 25804.3 | 28127.2 | 30539.0 | 24740.4
On average, L1 gives an RMSE of 28075.2 (std: 1151), while L2 gives an RMSE of 29176.4 (std: 4246.7). Not only is the L1 model better performing, but it is also more robust when it comes to handling data variations since its std is lower.