Widely accepted theory is that working memory can hold on average seven (7) plus minus two concepts at a time varying with each person. It is important to note the distinction between a concept and let's say a number or word. Concept can be anything, from a numbert to a word to an abstract meaning.
For example, the numbers 684804791 might feel overwhelmingly difficult to remember, but separing them to 684, 804 and 791 and assigning a (meaningful) mental image to each one, or an auditory tone for each (as the set of numbers is pronounced, the task could require just three slots in working memory instead of 12. Similarly a sentence, and its meaning, is more easier and useful to remember than all the invidual words, which just proves that memory doesn't have to work by just storing invidual letters, words or sentences. There's no limit to what kind of concepts it can hold.
Working memory holds concepts only a limited amount of time. To overcome this, a person might repeat a certain word or a phone number aloud, or mentally, until the task requiring the information is complete. Such as finding a piece of paper and scribbling down the important phone number.
Information from the working memory gets transferred to the long-term memory. The theory is, that it is not actually the amount of repeatition that triggers this process. In fact, the meaningfulness, and the connectity of the information is more important. In practical terms, continous repeation of foreign language words might be very inefficient. Instead, thinking of more creative ways to connect the foreign words to their native counterparts, such as mental images connecting the two words, is far more rewarding.
Arguments that WM is distinct from long-term memory come from: