Nous Research’s Arete-27B post-trains Qwen3.8-27B to reinforce Western values, arguing today’s strongest Chinese open models ship with state-aligned behavior in the weights; they score it on an 803-prompt eval across seven value categories including freedom of speech.