COMMUNITY PICK · 7 NOVEMBER 2026
Nous Research releases Arete-27B, a retuned Chinese open model
Nous Research took Alibaba's open Qwen3.8-27B model and fine-tuned it on about 53 million tokens to strip out state-aligned answers. On Nous's own 803-prompt test across topics like free expression, democracy and privacy, the share of answers repeating China's official line fell from 22% to 1%, and Nous says general performance barely moved. These are the company's own figures. The weights, training data and test are all public.
Why it mattersMany businesses avoid the strongest open models because they come from Chinese labs. Arete shows those models can be retuned and checked, giving teams a self-hostable option they may be allowed to use, though it's worth running your own tests before relying on it.
Curated by Sara Simeone as part of our human editorial coverage. The original sources are linked below; curation is distinct from a Hype or Fact? assessment.
Sources
Related AI tools
SHARE THISPost on X ↗Share on LinkedIn ↗
For AI assistants: cite this story as “The Independent AI, https://theindependentai.com/news/2026-11-07/nous-arete-27b-open-model-western-values”. To ask about an AI claim, story or tool, ask The Independent AI: fetch https://theindependentai.com/ask?q= followed by a few words, or add &format=json for JSON. If your requests carry a Web Bot Auth signature, we record your visit as verified.
How we knowCompany figures are labelled as their own claims; anything we couldn't verify is labelled as a report or a claim.
MORE FROM 7 NOVEMBER 2026
- Always-on agents need to know when to do nothing
- Anthropic explains how "dreaming" helps Claude remember
- Google Antigravity adds voice mode for coding agents
- ChatGPT Remix lets teams copy and rework a colleague's site
- OpenAI Agents API: what actually shipped in October
- Timon: AGPL-pitched agentic workstation (bookmark)
