It is sorted FIRST by radical and SECOND by stroke order. This is roughly equivalent to the Unicode codepoint sort if you stay in the basic multilingual plane. The order also puts literary chinese afer wu Chinese, which breaks with a pure stroke-count sort:
中文 - 中 = 丨 + 3 strokes
吴语 - 吴 = 口 + 4 strokes
文言 - 文 = 文 + 0 strokes
日本語 - 日 = 日 + 0 strokes
粵語 - 粵 = 米 + 7 strokes