BH provides a stricter control of the Type I error.

BH controls the False Discovery Rate (FDR) and is generally more powerful.

BH ensures that the probability of at least one false positive is exactly zero.

BH does not require the p-values to be independent.