The main difference between LSTM and GRU formulations $($classical$)$ are the following:

a$)$ the forget gate and input gate in LSTM are replaced by a single update gate

b$)$ the block output and cell state output are the same in GRUs, i$.$e$.$ there is no output gate

c$)$ the cell$-$value generative function is modified from tanh to sigmoid

d$)$ both $($a$)$ and $($b$)$ are true.