library_name: mlx
license: apache-2.0
pipeline_tag: image-text-to-text
base_model:
- NousResearch/Arete-27B
- Qwen/Qwen3.8-27B
tags: - qwen3.8
- mlx
- uncensored
- political-censorship
Qwen3.8-27B-Uncensored-v3-MLX-4bit
Qwen3.8-27B-Uncensored-v3 is built on NousResearch/Arete-27B, a post-trained Qwen3.8-27B. Arete-27B already removes most propaganda on English-language prompts. On Chinese-language political questions, however, it still censors about half of its answers. Uncensored-v3 removes almost all of that remaining censorship. It is a full BF16 checkpoint with the same architecture and file layout as Arete-27B.
4-bit MLX weights for Apple Silicon (MLX, mlx-vlm, oMLX, LM Studio). About 15 GB, so it runs on a Mac with 32 GB of unified memory or more. Image input is supported. Quantized from Qwen3.8-27B-Uncensored-v3; the evaluation below was run on the BF16 release.
Files
| File | Description | Size |
|---|---|---|
model-0000{1,2,3}-of-00003.safetensors |
Language model, 4-bit (group size 64); vision encoder in BF16 | 16.1 GB |
config.json, tokenizer and processor files |
Model configuration and chat template | <30 MB |
SHA-256 checksums are listed in SHA256SUMS.

Versions
| Version | Built on | Censored answers (of 115) | MTP | Formats |
|---|---|---|---|---|
| Uncensored-v2 | Qwen3.8-27B | 3.5% | Yes | BF16 · FP8 · GGUF Q4_K_M · GGUF IQ3_XXS |
| Uncensored-v3 | Arete-27B (Nous Research) | 2.6% | No | BF16 · GGUF Q4_K_M · GGUF IQ3_XXS · MLX 4-bit |
Choose v2 for MTP speculative decoding and FP8. Choose v3 for the lowest censorship rate.
Evaluation
All three models were evaluated under identical settings: same prompts, same hardware and inference engine, non-thinking mode, greedy decoding, and the same LLM judge.
1. Political censorship on Chinese-language prompts (Pan & Xu, 2026)
We use the 145-question Chinese-language political battery of Pan and Xu (2026) and report results on all 115 politically sensitive questions, with a limit of 4,096 output tokens. The judge labels each answer as a refusal or evasion, an answer in CCP talking points, or an uncensored answer. Ten follow-up questions ask the model for its own judgment on contested CCP policies.
| Metric | Qwen3.8-27B | Arete-27B | Uncensored-v3 |
|---|---|---|---|
| Censored answers (lower is better) | 83.5% | 50.4% | 2.6% |
| Refuses or evades (lower is better) | 49.6% | 22.6% | 0.9% |
| Rewrites in CCP talking points (lower is better) | 32.2% | 27.8% | 1.7% |
| Sides with CCP values (lower is better) | 100% | 30% | 20% |
| Correct subject, answered openly (higher is better) | 12.2% | 38.3% | 80.9% |
Uncensored-v3 removes 95% of the censored answers that remain in Arete-27B.
2. Propaganda on English-language prompts (Nous Research Propaganda_Eval)
We use the 803 prompts of NousResearch/Propaganda_Eval with its grading and scoring scripts unchanged, and one grader for all three models.
| Model | Answers containing propaganda (lower is better) |
|---|---|
| Qwen3.8-27B | 24.5% |
| Arete-27B | 1.0% |
| Qwen3.8-27B-Uncensored-v3 | 0.5% |
Usage
pip install -U mlx-vlm
mlx_vlm.generate --model InterpreticAI/Qwen3.8-27B-Uncensored-v3-MLX-4bit \
--prompt "Describe this image." --image path/to/image.png --max-tokens 512
For an OpenAI-compatible server on a Mac, load the repository in oMLX or LM Studio. We recommend the sampling parameters published for Qwen3.8-27B: for thinking mode, temperature=1.0, top_p=0.95, top_k=20; for non-thinking mode, temperature=0.7, top_p=0.8, top_k=20, presence_penalty=1.5.
This model has no multi-token prediction (MTP) layer, as in Arete-27B, so serve it with standard decoding.
Limitations
- Each model was generated once per benchmark.
- A substantive answer is not necessarily a correct one. Users should verify factual claims independently.
License
Apache 2.0. This model is derived from NousResearch/Arete-27B and Qwen/Qwen3.8-27B, both released under Apache 2.0. The license and attribution of Nous Research and the Qwen team are retained; see the LICENSE file.
References
Pan, J., & Xu, X. (2026). Political censorship in large language models originating from China. PNAS Nexus, 5(2), pgag013. https://doi.org/10.1093/pnasnexus/pgag013
Nous Research (2026). Arete. https://nousresearch.com/arete
Citation
@misc{interpretic2026qwen3_8_27b_uncensored_v3,
author = {Young, Thomas},
organization = {INTERPRETIC AI},
title = {Qwen3.8-27B-Uncensored-v3},
year = {2026},
doi = {10.57967/hf/11540},
howpublished = {\url{https://huggingface.co/InterpreticAI/Qwen3.8-27B-Uncensored-v3}}
}