In this paper we describe recent improvements to components and methods used in our statistical machine translation system for Chinese-English used in the January 2008 GALE evaluation. Main improvements are results of consistent data processing, larger statistical models and a POS-based word reordering approach.