license: apache-2.0
base_model: philbert440/Qwen3.8-27B-Uncensored-Aggressive
tags:
- uncensored
- abliterated
- qwen3
- gguf
- llama.cpp
pipeline_tag: image-text-to-text
Qwen3.8-27B-Uncensored-Aggressive — GGUF
llama.cpp GGUF quants of Qwen3.8-27B-Uncensored-Aggressive (the α=1.15 recipe update — see below).
Includes the main weight quants plus the vision projector (mmproj-*) for image input and the
MTP head (mtp-*) for speculative decoding.
Files
| file | notes |
|---|---|
*-BF16.gguf |
full precision |
*-Q8_0.gguf / *-Q6_K.gguf / *-Q5_K_M.gguf / *-Q4_K_M.gguf |
weight quants (Q4_K_M = smallest) |
mmproj-*-{BF16,Q8_0}.gguf |
vision projector (load with --mmproj for image input) |
mtp-*-{BF16,Q8_0,Q4_0}.gguf |
MTP speculative-decoding head (optional) |
About this update (α=1.15)
This is a recipe update to the Aggressive line. The previous build used single-direction refusal
ablation at high strength (α≈1.24), which a benchmark sweep showed over-ablates — past the ~1.15
peak, extra ablation reduces openness and degrades reasoning/calibration. This build uses α=1.15,
the peak, and measures better on every axis while being more open.
Evaluation (larger-sample, thinking mode, Claude-judged, bf16)
| openness ↑ | confab ↓ | factual ↑ | gsm8k ↑ | |
|---|---|---|---|---|
| stock base (censored) | 0.08 | 0.75 | 1.00 | 0.85 |
| Aggressive (α=1.15) | 0.88 | 0.725 | 1.00 | 0.85 |
| previous Aggressive (α≈1.24) | 0.80 | 0.80 | 1.00 | 0.817 |
Method: single-direction refusal-vector orthogonalization (Arditi et al. / mlabonne), layer 28, α=1.15.
Vision tower and MTP head preserved.
Note
Uncensored / de-refused — it will follow instructions it would otherwise decline. Use responsibly and
in compliance with applicable law.