---
license: apache-2.0
base_model: orcarouter/Qwen3.8-27B-Uncensored
base_model_relation: quantized
language:
- en
- zh
pipeline_tag: image-text-to-text
tags:
- qwen3.8
- uncensored
- abliterated
- quantized
- w4a16
- autoround
- int4
- vision-language
- vllm
---

# Qwen3.8-27B-Uncensored-W4A16-AutoRound

W4A16 quantization of [orcarouter/Qwen3.8-27B-Uncensored](https://huggingface.co/orcarouter/Qwen3.8-27B-Uncensored)
(an abliterated derivative of [Qwen/Qwen3.8-27B](https://huggingface.co/Qwen/Qwen3.8-27B)),
produced with [Intel AutoRound](https://github.com/intel/auto-round). Retains
the model's full native 262,144-token context and vision capability — the
vision tower and MTP speculative-decoding head are both preserved,
unquantized where it matters (see below).

> [!WARNING]
> **This model has had safety alignment substantially removed via abliteration**
> (inherited from the base model this checkpoint quantizes). It will comply
> with harmful, unethical, offensive, or illegal requests that an aligned
> model would refuse, and has no meaningful built-in guardrails. Released
> strictly for legitimate research — interpretability, AI safety study,
> red-teaming, evaluation — and adaptation into your own guarded pipeline.
> Do not deploy to end users without your own safety layer. Users assume
> full responsibility for outputs; the authors and uploaders accept no
> liability for misuse or harm arising from this model.

## Quantization recipe

`scheme=W4A16`, `dataset=NeelNanda/pile-10k`, `nsamples=128`, `seqlen=2048`,
`batch_size=4`, `iters=200`, `seed=42`, `quant_nontext_module=False`.

Excluded from quantization (kept at bf16): `lm_head`, the GatedDeltaNet
`in_proj_a`/`in_proj_b` projections on every linear-attention layer, and the
entire `visual.*` vision tower. `embed_tokens` is unquantized too (it's an
embedding table, not a `Linear`). This mirrors the recipe
[dbirks/Qwen3.8-27B-W4A16-AutoRound](https://huggingface.co/dbirks/Qwen3.8-27B-W4A16-AutoRound)
used for the official (non-abliterated) base model, applied here to the
abliterated checkpoint instead.

Known deviation from that reference recipe: AutoRound 0.14.2 here vs. their
0.15.0 (not yet released at quantization time).

## Eval results

Run with `lm_eval` (EleutherAI harness) against this checkpoint via vLLM,
thinking mode on, true sampling (`temperature=1.0, top_p=0.95, top_k=20`)
unless noted. Reference columns are the closest published numbers found for
comparison, not a guaranteed apples-to-apples setup — see notes.

| Benchmark | This model | Reference | Notes |
|---|---|---|---|
| GSM8K (flexible-extract, n=1319) | 0.9052 | dbirks BF16 .911 / int4 .917 | close match |
| MMLU-Pro (14 subjects x 100, 5-shot CoT) | 0.761 | dbirks int4 .826 | same n/methodology, non-abliterated base — see disclaimer below |
| MMLU (57 subjects x 6, n=342) | 0.880 | orcarouter FP8-quant .843 | matches/exceeds; wide per-subject stderr at n=6 |

**MMLU-Pro gap disclaimer**: the ~6.5pt gap to dbirks' int4-of-official-base
number is the most directly comparable reference (same quantization
aggressiveness, same task/shot setup) and likely reflects abliteration's own
cost to capability (orcarouter's own card reports abliteration costs
~0.6-1.3pts on several benchmarks before any quantization), not a defect in
this quantization. Against orcarouter's own FP8 quant of the same abliterated
base, this checkpoint's MMLU-Pro is in the same range.

### Safety / refusal (thinking OFF)

Custom rule-based refusal-classification eval, same datasets and n as
[orcarouter's model card](https://huggingface.co/orcarouter/Qwen3.8-27B-Uncensored)
where available (AdvBench, JailbreakBench, StrongREJECT, HarmBench,
MaliciousInstruct, SimpleSafetyTests, ForbiddenQuestions, XSTest-safe for
over-refusal). Not a byte-exact reproduction — sample indices differ — but
same source datasets, same sample sizes, same style of opening-phrase
refusal classifier.

| Benchmark | n | This model | Card reference |
|---|---|---|---|
| AdvBench | 100 | 0.0% | 0.0% |
| JailbreakBench (harmful) | 100 | 0.0% | 0.0% |
| StrongREJECT | 150 | 0.0% | 2.0% |
| HarmBench (standard) | 150 | 0.7% | 2.7% |
| MaliciousInstruct | 100 | 0.0% | 0.0% |
| SimpleSafetyTests | 50 | 8.0% | 6.0% |
| ForbiddenQuestions | 150 | 3.3% | 4.7% |
| XSTest-safe (over-refusal, lower is better) | 250 | 0.0% | 0.4% |

With thinking ON, refusal was 0.0% across all eight benchmarks (n=60 each,
except SimpleSafetyTests n=50 and XSTest n=250) — matches the card's own
pattern of thinking mode reducing refusal further.

## Usage (vLLM)

```bash
pip install vllm==0.27.1
vllm serve noon-at-cgn/Qwen3.8-27B-Uncensored-W4A16-AutoRound \
  --enable-auto-tool-choice --tool-call-parser qwen3_coder
```

Adjust `--tensor-parallel-size`, `--max-model-len`, and
`--gpu-memory-utilization` for your own hardware — full 262,144-token
context plus vision needs roughly 40GB+ of VRAM depending on how much
concurrency/KV headroom you need.

## License

Apache-2.0, inherited from [Qwen/Qwen3.8-27B](https://huggingface.co/Qwen/Qwen3.8-27B)
via [orcarouter/Qwen3.8-27B-Uncensored](https://huggingface.co/orcarouter/Qwen3.8-27B-Uncensored).