L1 regularizations penalize the model for having a parameter, while L2 penalizes the model for having a parameter with a large magnitude.
L1 penalizes variables more than L2 which squares errors thereby seeing higher error and minimizing that squared error.