Rank Moore language models based on their performance across translation, speech, and understanding tasks