If leaky bucket means "manages burst sizes and burst intervals" then yeah, they are similar in that regard but they do it differently. GCRA is more analogous to a traffic light where its solely time-based and there is no overhead needed to remove/add tokens so you are storing 1 less state object. For all purposes its probably a micro-optimization but being aware how GCRA works can yield less code but perhaps at the expense of obscurity of a lesser known algo.

I think you understand the differences but I think its improper to consider it a variant of leaky bucket as various resources refute