It requires estimating the parameter under the alternative hypothesis.

It only requires evaluating the derivative of the log-likelihood at the null hypothesis value $\theta_0$.

It is always more powerful than the Wald test in small samples.

It is the same as the likelihood ratio test for all models.