get_usage() does not give exact character count

[jaixan]

Hi,

If I call translator.get_usage() before and after a translation, I get the same number of character used, an even 1 000 000. Is there a way to get the exact character count?