Nous Research releases Arete-27B, a Qwen model post-trained to remove CCP-aligned answers

Nous Research has released Arete-27B, a version of Alibaba's open-weight Qwen3.8-27B that it post-trained "to reinforce Western values on a Chinese-trained model," according to a Nous Research blog post. The company said it observed refusals, deflections and CCP-framed answers on politically sensitive topics in the base model.

On Nous Research's own evaluation of 803 prompts across seven value categories, including freedom of expression, state ideology, democracy and data privacy, Arete-27B raised its fact score from 2.32 to 2.98 out of 5 and cut answers containing propaganda from 22% to 0.9%, the model card says. The improvement held in all seven categories without a meaningful regression on the general-capability and agentic benchmarks the team ran.

The model was trained with supervised fine-tuning only, with no reinforcement learning, preference optimisation or architectural changes. Nous Research said it does not claim to have removed every behaviour tied to the base model's viewpoint, and argued that open-weight models let labs adapt the values a model ships with. The US government has taken steps to restrict Chinese AI models, especially DeepSeek, from federal use, the company noted.