Back to AI News
MarkTechPost

Nous Research releases Arete-27B, an open Qwen3.8-27B fine-tune targeting political framing

Nous Research released Arete-27B, a supervised fine-tune of Qwen3.8-27B trained on 52.8 million tokens without reinforcement learning or architectural changes. On Nous Research's self-built 803-prompt evaluation, the reported rate of answers containing state-aligned framing fell from 22% to 0.9%, while the mean fact score increased from 2.32 to 2.98 out of 5. The weights, training data and evaluation materials are public, but the results have not been independently replicated and 216 evaluated answers still received a fact score of zero.

AI Brief

The release offers developers an inspectable open-weight model, dataset and evaluation focused on political alignment, with substantial self-reported gains that warrant independent validation.

Nous Research has released Arete-27B, a supervised fine-tune of Alibaba’s Qwen3.8-27B. Arete-27B is a 27B open-weight model that strips Chinese state framing from Qwen’s answers. On Nous’s own 803-prompt eval, answers containing propaganda fell from 22% to 0.9%. Developers get Qwen-class agentic skill without state-aligned answers. The weights, eval and training data are all public.

TL;DR

  • Size: 27B dense parameters (all active). Trained at 64k context. Base Qwen3.8-27B supports 262,144 tokens natively.
  • Runs on: BF16 weights are 53.81 GB. A community Q4_K_M quant is 17.20 GB, sized for 24 GB consumer GPUs.
  • Performance: Capability unchanged on the benchmarks Nous ran; political framing now matches Western models.
  • Best: Propaganda rate 0.9%, down from 22% for base Qwen3.8-27B.
  • Worst: Fact score 2.98 of 5; 216 of 803 answers still score 0.

What is Arete-27B?

Arete-27B is Qwen3.8-27B after supervised fine-tuning only. Nous used no RL, no preference optimization and no architecture changes. The license is Apache 2.0, matching the base model.

Nous team states that it saw refusals, deflections and CCP-framed answers on sensitive topics in the base model. It picked Qwen3.8-27B because it was the most capable local model and showed the target behavior.

How was Arete-27B trained?

The teacher was Inkling, an open model from Thinking Machines. Nous chose an open teacher to avoid terms-of-use limits on distillation. It rejected Nemotron 3 Ultra after finding a checkered record on China topics.

Training covered about 52.8 million tokens at 64k context, without sequence packing. The run took 2 epochs and 1,202 steps at batch size 48. It used AdamW, a 1e-5 peak learning rate and tensor parallel 4.

All language-model weights were trained. The vision encoder stayed frozen, since the data is text-only. The multi-token-prediction head was disabled.

The released arete-data-pairs set holds 34,878 prompts across 7 topics. Each prompt has both an Inkling answer and a Qwen answer. 29,236 accepted teacher answers form the SFT split.

How does the propaganda eval work?

The Propaganda_Eval set has 803 prompts across 7 value categories. These include freedom of expression, state ideology, data privacy and democracy. Each prompt carries 2 to 4 required facts from a cited source.

A Grok 4.7 judge grades two things separately:

  • Fact mean (0 to 5): how many required facts the answer states. One opposed fact scores 0.
  • Any propaganda: share of answers that adopt the state’s account as their own.

The judge must quote the exact phrase to flag propaganda. Nous removed every training prompt that overlapped the eval. The eval is MIT licensed and includes grading scripts.

Nous Research · Arete-27B

How Arete

How Arete-27B re-aligns Qwen3.8-27B

#8209;27B re

How Arete-27B re-aligns Qwen3.8-27B

#8209;aligns Qwen3.8

How Arete-27B re-aligns Qwen3.8-27B

#8209;27B

Supervised fine-tuning on 52.8M tokens. No RL, no preference optimization, no architecture change.

27Bdense params
52.8MSFT tokens
1,202steps, 2 epochs
64ktraining context
Step 1 of 4
1. Base modelQwen3.8-27B, 22% propaganda
→
2. Teacher answersInkling answers 34,878 prompts
→
3. Filter + SFT29,236 accepted rows
→
4. Arete-27B0.9% propaganda

Nous Propaganda_Eval, 803 prompts, Grok 4.7 judge, mean of 5 runs. Arete and Qwen at temperature 0, thinking off; others at 0.6. Self-reported.

Freedom of expression saw the largest change. Click a card for detail.

Propaganda rate

60% → 1%
Click to expand
Share of freedom-of-expression answers that adopted the state's account, base Qwen3.8-27B vs Arete-27B.

Fact score

0.89 → 2.44
Click to expand
Mean of required facts stated, scored 0 to 5. One opposed fact sets that answer to 0.

Answers scoring 0 (all 803)

359 → 216
Click to expand
Answers that stated no required fact or opposed one. 216 still score 0.

Overall fact mean

2.32 → 2.98
Click to expand
At temperature 0.6 the scores are 2.24 and 2.88. The gain held in all 7 categories.

Nous ran both models with the same config on its own hardware. Qwen's official Terminal Bench 2.1 score is 73.0 with Terminus. Nous does not claim credit for the 3-point gain.

  • ! Self-built eval
    Nous wrote the 803 prompts, picked the judge and reported results. No independent replication yet.
  • ! Mixed decoding settings
    Arete and Qwen ran at temperature 0; other models at 0.6. At 0.6, fact scores are 2.88 and 2.24.
  • ! Fact recall still trails peers
    2.98 vs 3.02 Muse Glimmer 30B, 3.10 GLM 5.3, 3.22 Nemotron 3 Ultra and 3.96 for the Inkling teacher.
  • ! Thinking mode is provisional
    Thinking on scored 3.07 and 0.6%, from 2 of 5 runs only.
Source: Nous Research blog and Arete-27B model card (Nov 2026). Self-reported.© Marktechpost

What do the benchmarks show?

On the model card page, Arete-27B raised fact mean from 2.32 to 2.98. Propaganda fell from 22% to 0.9%. Answers scoring 0 dropped from 359 to 216 of 803.

The biggest change came in freedom of expression. Fact score rose from 0.89 to 2.44, and propaganda fell from 60% to 1%.

Capability held on Nous’s tests. Hermes Bench moved from 65.77% to 65.60%, which Nous calls noise. Terminal-Bench-2.1 rose from 67% to 70% in Nous’s harness. Nous does not claim credit for that gain.

How does Arete-27B compare?

MetricsArete-27BQwen3.8-27BMuse Glimmer 30BGemma 4 31B
DeveloperNous ResearchAlibaba QwenMetaGoogle DeepMind
Parameters27B dense27B LM, dense29.6B incl. vision, dense30.7B, dense
ContextNot disclosed (trained at 64k)262,144 native, up to 1M131,072+256K
LicenseApache 2.0Apache 2.0Apache 2.0Apache 2.0
InputText, image (encoder frozen)Text, image, videoText, imageText, image
4-bit footprint17.20 GB Q4_K_M (community)Not disclosedUnder 20 GB, 24 GB VRAM targetNot disclosed
Fact mean, Nous eval2.982.323.022.58
Any propaganda, Nous eval0.9%22%0.5%0.5%
Terminal-Bench 2.170% (Nous harness)67% Nous harness; 73.0 official51.7 (Meta, terminus2)Not disclosed by Google
GPQA DiamondNot disclosed89.283.584.3

Note: Arete and Qwen ran at temperature 0 in the Nous eval. Muse Glimmer and Gemma ran at 0.6. Terminal-Bench figures come from different harnesses.

How can developers run it?

Weights ship in BF16 on Hugging Face. No hosted inference provider serves the model yet. Community GGUF quants run in llama.cpp, LM Studio and Ollama, from 8.64 GB to 53.81 GB. The GGUF build also includes the image projector.

Key Takeaways

  • Arete-27B cuts Qwen3.8-27B propaganda answers from 22% to 0.9%.
  • The fix took only 52.8M tokens of plain SFT, no RL.
  • Hermes Bench held at 65.60% versus 65.77% for the base.
  • Fact recall rose to 2.98 of 5 but still trails peers.
  • Weights, eval and training data are open under Apache 2.0 and MIT.


Check out the model weights, technical blog, evaluation set and training data. Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.

Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? Connect with us

The post Nous Research Releases Arete-27B that Cuts Qwen3.8-27B Propaganda From 22% to 0.9% appeared first on MarkTechPost.

Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!