We test two popular methods to run local language models on your Linux box. Then, we push the limits to see which language models will toe the line and which won't.