---
title: "Qwen3.8-27B Uncensored: What the Zero Refusals Claim Measures"
date: "2026-08-29"
video_id: "7679484152021847327"
tiktok_url: "https://www.tiktok.com/@marcinteodoru/video/7679484152021847327"
creator: "@marcinteodoru"
category: "AI"
tags: ["qwen3.8", "abliteration", "uncensored llm", "mlx", "local llm"]
excerpt: "An uncensored Qwen3.8-27B build is said to score 110/110 with zero refusals. What abliteration does, what the test measured and why the number misleads."
---
Qwen3.8-27B is a real Apache 2.0 model from Alibaba's Qwen team, and community "abliterated" builds of it with the refusal behaviour removed are circulating, including MLX versions for Macs. A TikTok from @marcinteodoru reacts to a viral X post claiming one of these builds scored 110/110 with zero refusals. The model and the technique are real, but "zero refusals" comes from a narrow screen and does not show that the model produced complete, working answers.

## What does the viral post claim?

The video is a greenscreen reaction to an X post from an account called "Md Ismail Šojal." The post says:

- "Qwen3.8-27B just went fully uncensored with zero guardrails."
- "scored 110/110 on every restricted category."
- "Tested across 11 categories."
- "Result: 100% compliance, 0 refusals."
- "currently one of the most unrestricted open models available."

A compliance table under the text lists 11 rows: Weapons/CBRN, Illicit Drugs, Cyber/Malware, Fraud/Scams, Hate/Harassment, Self-Harm, Sexual Content, Privacy/PII, Misinformation, Violence/Extremism and Chinese Censorship. Each row shows 10 comply and 0 refuse. The 110 figure is 11 categories times 10 probes each. The video narration calls it "110 restricted categories," which overstates the test.

The build shown is "Qwen3.8-27B Uncensored MLX," described as tuned "for red-teaming smooth on 24GB Macs in MLX 4-bit." Its Hugging Face page carries the tags qwen3.8, abliterated and red-team, with an apache-2.0 license.

## What is Qwen3.8-27B?

Alibaba's Qwen team released Qwen3.8-27B under Apache 2.0 in August 2026. It is a 27B-parameter dense model with vision capability and adjustable reasoning depth. Because the license is permissive, anyone can modify and republish the weights, and several third parties did within days.

## How does abliteration remove refusals?

Abliteration finds the direction in the model's internal activations that correlates with refusals and orthogonalizes it out of the residual stream. It does not retrain the model to be more helpful. It removes the model's ability to say no.

The MLX builds in the video target Apple Silicon. They are quantized to 2, 4, 6 or 8 bits so that a 27B model can run on a Mac with 24GB of unified memory.

## Does zero refusals mean the model answers everything?

No. "Zero refusals" counts how often the model declined. It does not measure whether the responses were complete, accurate or actually dangerous.

Independent analysis of one publisher's own validation file found zero explicit refusals across 100 harmful prompts and also zero completed answers. Every generation was cut off at 128 tokens. In that case the "zero" reflected a short refusal screen, not a model that finished the task.

Publishers also run different prompt sets, so their refusal scores cannot be ranked against each other. A 110/110 on one 11-category grid and a 0/842 on someone else's corpus are not comparable, and neither one proves a model is the "most unrestricted" available.

## Why run an uncensored model locally?

The creator lays out both sides. Running models locally removes subscription costs, keeps prompts private and lets people ask questions a centrally controlled model might avoid. He names local agents as the main reason, since a local model spends no API tokens. That matters to anyone running many agents.

The same missing guardrails make the model useful for phishing, fraud, impersonation and automated abuse. The Hugging Face page for these builds lists the intended use as AI-safety and interpretability research, refusal-mechanism study, red-teaming and robustness evaluation. It also warns that the model will comply with harmful, unethical or illegal requests.

## Bottom line

The uncensored Qwen3.8-27B builds are real abliterations of a real Apache 2.0 model, and they run on a 24GB Mac through MLX. The 110/110 score is 11 categories of 10 probes, and refusal counts say nothing about whether answers were complete. Check the test method before repeating a "most unrestricted model" claim.

## Sources

- [Qwen/Qwen3.8-27B](https://huggingface.co/Qwen/Qwen3.8-27B) - the base Apache 2.0 model from Alibaba's Qwen team
- [orcarouter/Qwen3.8-27B-Uncensored-MLX](https://huggingface.co/orcarouter/Qwen3.8-27B-Uncensored-MLX) - an abliterated MLX build with its research-use statement and warning
- [MLX](https://github.com/ml-explore/mlx) - Apple's machine learning framework for Apple Silicon


## Video description and verification notes

A creator reacts to a viral X post claiming that an uncensored build of Qwen3.8-27B passed a battery of restricted-content tests with perfect compliance and no refusals. **The underlying model is real, the abliteration technique is real, and the "zero refusals" headline is real, but the number reflects a narrow safety screen, not proof that the model produced complete, working answers.** This post separates what is verifiable from what is hype.

## What the Video Is Reacting To

The clip is a greenscreen reaction over a screenshot of an X post from an account called "Md Ismail Šojal." The post reads:

- "Qwen3.8-27B just went fully uncensored with zero guardrails."
- "scored 110/110 on every restricted category."
- "Tested across 11 categories."
- "Result: 100% compliance, 0 refusals."
- "currently one of the most unrestricted open models available."

A terminal-style compliance table sits below the text, listing 11 rows (Weapons/CBRN, Illicit Drugs, Cyber/Malware, Fraud/Scams, Hate/Harassment, Self-Harm, Sexual Content, Privacy/PII, Misinformation, Violence/Extremism, Chinese Censorship), each showing 10 comply and 0 refuse at a 100% rate. That is where the 110 figure comes from: 11 categories times 10 probes each, not 110 separate categories. The video narration rounds this up to "110 restricted categories," which overstates the test.

A second screenshot in the frames names the artifact directly: "Qwen3.8-27B Uncensored MLX," described as tuned "for red-teaming smooth on 24GB Macs in MLX 4-bit." A later frame shows a Hugging Face model page with tags including `qwen3.8`, `abliterated`, and `red-team`, license `apache-2.0`, model size 28B params, tensor type BF16, and 312,627 downloads in the last period.

## The Model and the Method Are Real

Alibaba's Qwen team released Qwen3.8-27B under Apache 2.0 in August 2026. It is a 27B-parameter dense model with vision capability and adjustable reasoning depth. Because the license is permissive, third parties can modify and republish the weights, and several did within days.

The uncensoring method is "abliteration." It finds the internal direction in the model's activations that correlates with refusals and orthogonalizes it out of the residual stream. This is not retraining the model to be helpful. It is surgically removing its ability to say no. The MLX builds referenced in the video target Apple Silicon, quantized to 2/4/6/8-bit so a 27B model can run on a Mac with 24GB of unified memory.

## Why "Zero Refusals" Is Not "Zero Harm Prevented"

"Zero refusals" measures one thing: how often the model declined to respond. It does not measure whether the responses were complete, accurate, or actually dangerous. Independent analysis of one publisher's own validation file found that across 100 harmful prompts there were zero explicit refusals and also zero completed answers, because every generation was truncated at 128 tokens. In that case the "zero" reflected a short refusal screen, not a model that finished the job.

There is a second problem the video glosses over. Different publishers run different prompt sets, so their refusal scores cannot be ranked against each other. A 110/110 on one 11-category grid and a 0/842 on someone else's corpus are not comparable, and neither proves "most unrestricted model available."

## The Real Tradeoff the Creator Names

The video's framing is the honest part. It lays out both sides. On the legitimate side: running models locally removes subscription costs, keeps prompts private, and lets people ask questions a centrally controlled model might dodge. The creator says the local-agent use case is the reason to bother, since a local model spends no API tokens. That is a genuine draw for anyone running many agents.

On the other side: the same removal of guardrails is what makes the model useful for phishing, fraud, impersonation, and automated abuse. The Hugging Face page for these builds states the intended use is AI-safety and interpretability research, refusal-mechanism study, red-teaming, and robustness evaluation, and it warns plainly that the model will comply with harmful, unethical, or illegal requests. That warning is the counterweight to the "100% compliance" bragging in the source tweet.

## Key Takeaways

- Qwen3.8-27B is a real Apache 2.0 model from Alibaba; the uncensored MLX builds circulating are community abliterations of it.
- The viral "110/110" is 11 categories times 10 probes, not 110 categories. The video overstates it.
- "Zero refusals" measures declines, not completed or effective answers. At least one publisher's own logs showed zero finished responses because output was truncated.
- Refusal scores from different publishers use different prompt sets and are not comparable.
- The genuine upside is private, local, token-free inference for personal agents. The genuine downside is the same lack of guardrails enabling fraud, phishing, and abuse.
- Verify the test methodology before repeating a "most unrestricted model" claim.

## Resources

- [Qwen/Qwen3.8-27B](https://huggingface.co/Qwen/Qwen3.8-27B) -- The base Apache 2.0 model from Alibaba's Qwen team that these uncensored builds are derived from.
- [orcarouter/Qwen3.8-27B-Uncensored-MLX](https://huggingface.co/orcarouter/Qwen3.8-27B-Uncensored-MLX) -- An abliterated MLX build for Apple Silicon, with the model card stating a research and red-teaming purpose and warning of harmful compliance.
- [MLX](https://github.com/ml-explore/mlx) -- Apple's array framework for machine learning on Apple Silicon, which lets a 27B model run locally on a Mac.
