In our lab we often measure working memory as a background measure to other tasks. With primary school age children we use the Corsi-block test to measure visuo-spatial working memory and a Digit span test to measure verbal working memory (often we use a backwards version of this task as it can be more sensitive to developmental differences). For both of these tasks the child is presented with a sequence which they have to repeat. The length of the sequence increases until children are unable to do the task anymore. So, with the Corsi-block task, the experimenter taps out a sequence across a number of blocks (up to nine blocks) for the child to repeat, and on the digit span task the experimenter verbalises a sequence of numbers which the child has to repeat (forwards or backwards, depending on the task). Yes, the Baddeley model is still useful, particularly with reference to the slave systems (verbal and visuo-spatial working memory).
I have also used the Corsi-block and Digit Span tasks in young primary school-aged children. However, I have found that children younger than 5 really struggle with these tasks, and particularly the Backwards Digit Span. A similar task to the Corsi-block test is the Dot Matrix test where children have to remember a series of locations of a dot in a 3 by 3 matrix.