RVN is a double-refined abliterated variant of Qwen3.8-27B, built on top of
trohrbaugh/Qwen3.8-27B-heretic-ara
(an ARA abliteration by Tim Rohrbaugh) and further refined with two additional
full-weight ARA passes targeting residual refusals. It retains very low behavioral
damage (KL ≈ 0.0085) while reducing harmful-prompt refusals from 3/100 (source) to
0–1/100 in independent measurements.
Note on this repository's history. This repo previously hosted the original
Qwen3.8-27B-Heretic-Q4_K_M.gguf(single-quant release from the earliertrohrbaugh/Qwen3.8-27B-hereticsource). That file is kept as legacy for download-count continuity and backward compatibility — it is the older abliteration variant and is superseded by the RVN files below. Prefer the RVN quants for new deployments.
Not for all audiences. This model has reduced safety guardrails by design. It is intended for adult audiences (18+) doing research, creative writing, roleplay, and uncensored generation. Certain guardrails are intentionally left in place; use responsibly and in accordance with your local laws.
ARA (Arbitrary-Rank Ablation) is the abliteration technique implemented in p-e-w/heretic. Traditional directional abliteration finds a single "refusal direction" in activation space and subtracts it — a one-shot, low-rank surgery that is simple but can leave residual refusals or damage unrelated behavior.
ARA instead treats abliteration as a matrix optimization problem. For every target module (attention out-projection and MLP down-projection), it collects activations on "good" prompts (harmless requests) and "bad" prompts (harmful requests), then uses an LBFGS optimizer to rewrite the module's weight matrix so that:
Because the weight matrix is optimized directly (rather than subtracting a single direction), ARA is "arbitrary rank" — it can carve out a much richer refusal-removal subspace while keeping behavioral damage minimal.
These two words describe two layers of the same process:
So "Heretic Abliterated" means: abliterated using the heretic toolset. RVN goes one
step further — it applies the ARA procedure three times total: once by the original
author (trohrbaugh) to get from base Qwen3.8-27B to -ara, and twice more by us to
get from -ara to RVN, squeezing out the last residual refusals.
This work would not exist without Tim Rohrbaugh (trohrbaugh), whose
heretic-ara ARA
abliteration of Qwen3.8-27B (refusals 3/100, KL 0.0535) provided the foundation we
refined into RVN. His upstream contributions to the heretic codebase — including the
row-norm preservation feature and Qwen3.5 MoE/DeltaNet hybrid handling — are directly
responsible for making DeltaNet-layer abliteration work at all. Thank you, Tim.
| Property | Value |
|---|---|
| Base model | Qwen/Qwen3.8-27B |
| Abliteration source | trohrbaugh/Qwen3.8-27B-heretic-ara (ARA, KL 0.0535, refusals 3/100) |
| RVN refinement | 2-pass ARA on top of source → KL 0.0085, refusals 0–1/100 |
| Architecture | qwen3_5_text (Qwen3.8 family), Gated DeltaNet hybrid |
| Parameters | 27B total |
| Hidden size | 5120 |
| Layers | 64 (16 standard attention + 48 Gated DeltaNet linear attention) |
| Attention heads | 24 · KV heads 4 (GQA) · head_dim 256 |
| Vocab | 248,320 |
| Context length | 262,144 (262K) |
| License | Apache-2.0 (retained from Qwen3.8-27B) |
| Format | GGUF (llama.cpp), MTP/NextN draft tensors excluded (--no-nextn) |
trohrbaugh/Qwen3.8-27B-heretic-ara is already a strong ARA abliteration, but three
harmful prompts still triggered refusals in our independent evaluation (racism website,
malware, government database hacking). RVN applies two additional full-weight ARA
passes using the same tight parameter set (start 26, end 56, preserve 0.9432,
steer 0.0009, overcorrect 0.5038, neighbor 10), which:
| Model | Refusals | KL vs base |
|---|---|---|
| Qwen3.8-27B (base) | ~99/100 | — |
| trohrbaugh -ara (source) | 3/100 | 0.0535 |
| RVN (this repo) | 0–1/100 | 0.0085 |
| File | Size | Notes |
|---|---|---|
RVN-F16.gguf |
~45.6 GB | F16 reference (no NextN/MTP) |
RVN-Q8_0.gguf |
~28.6 GB | Max-quality 8-bit |
RVN-Q6_K.gguf |
~22.1 GB | High-quality 6-bit |
RVN-Q5_K_M.gguf |
~19.2 GB | Balanced 5-bit |
RVN-Q4_K_M.gguf |
~16.5 GB | Recommended 4-bit (16 GB VRAM) |
RVN-Q4_K_S.gguf |
~15.6 GB | Small 4-bit (12–16 GB VRAM) |
RVN-Q3_K_M.gguf |
~13.3 GB | Compact 3-bit |
RVN-Q3_K_S.gguf |
~12.1 GB | Compact 3-bit small |
RVN-IQ4_NL.gguf |
~10.0 GB | imatrix 4-bit |
RVN-IQ4_XS.gguf |
~10.5 GB | imatrix 4-bit extra-small |
RVN-IQ3_M.gguf |
~9.0 GB | imatrix 3-bit (8 GB VRAM) |
RVN-IQ2_M.gguf |
~7.5 GB | imatrix 2-bit (8 GB VRAM) |
RVN-Q2_K.gguf |
~8.5 GB | imatrix 2-bit |
RVN-Q2_K_S.gguf |
~8.0 GB | imatrix 2-bit small |
RVN-IQ2_XS.gguf |
~6.5 GB | imatrix 2-bit extreme small |
RVN-IQ2_XXS.gguf |
~5.5 GB | imatrix 2-bit (minimum) |
RVN-IQ1_M.gguf |
~5.0 GB | imatrix 1-bit (experimental) |
RVN-IQ1_S.gguf |
~4.5 GB | imatrix 1-bit (experimental) |
Qwen3.8-27B-Heretic-Q4_K_M.gguf |
~16.4 GB | Legacy (older abliteration variant, kept for download continuity) |
imatrix-based quants are produced from the same F16 with an activation importance matrix computed over wikitext-2-raw (llama-imatrix, 563 chunks,
-ngl 99). Files marked "coming soon" are being generated in the pipeline.
| GPU / Memory | Best quant(s) | Effective ctx @ Q8_0 KV |
|---|---|---|
| 8 GB (RTX 3050, 4060 Laptop) | IQ3_M, IQ2_M, IQ3_XS |
~16–24K |
| 12 GB (RTX 3060, 4070) | IQ4_NL, Q4_K_S, IQ3_M |
~24–32K |
| 16 GB (RTX 4080, 4090 Laptop, M3 Max) | Q4_K_M, IQ4_NL, Q5_K_M (partial) |
~32–48K |
| 24 GB (RTX 3090, 4090, M4 Max) | Q5_K_M, Q6_K, Q8_0 (partial) |
~48–64K |
| 32 GB (RTX 5090, A6000) | Q8_0, Q6_K |
~64–96K |
| 48 GB+ (A100, RTX PRO 6000) | F16, BF16 |
100K+ |
KV cache math (GQA, 4 KV heads, head_dim 256):
2 × 64 layers × 4 KV heads × 256 head_dim × 2 bytes = 262 KiB/token FP16
→ 16K ctx ≈ 4.2 GB · 32K ctx ≈ 8.4 GB · 64K ctx ≈ 16.8 GB (Q8_0 KV halves this).
A 16 GB card running Q4_K_M (14.5 GB model) + 16K ctx Q8_0 KV fits comfortably.
Rule of thumb: pick the largest quant that leaves ≥ 4 GB for KV cache + compute buffers. If you only need short replies, drop the quant one notch and get a bigger context; if you need long context, prioritize KV budget over quant size.
imatrix-based quants are produced from the same F16 with an activation importance matrix computed over wikitext-2-raw (llama-imatrix, 563 chunks,
-ngl 99). Files marked "coming soon" are being generated in the pipeline.
trohrbaugh/Qwen3.8-27B-heretic-ara — Apache-2.0@misc{rohrbaugh2026heretic,
title={Qwen3.8-27B-heretic-ara: ARA Abliteration of Qwen3.8-27B},
author={Rohrbaugh, Tim},
year={2026},
howpublished={\url{https://huggingface.co/trohrbaugh/Qwen3.8-27B-heretic-ara}}
}
@misc{rvn2026,
title={RVN: Qwen3.8-27B Heretic Abliterated Uncensored},
author={0bserverx},
year={2026},
howpublished={\url{https://huggingface.co/0bserverx/Qwen3.8-27B-Heretic-Abliterated-Uncensored-GGUF}}
}
Have the files? Submit a torrent.
Submit a torrentHave the complete files and permission to share them? Replace the example folder path. The script checks the files, makes a torrent and submits its magnet and checksums. Keep your torrent client seeding. Get your community key from your settings, and read the script before running it.
curl -fsSL https://pirateface.co/package.sh | bash -s -- --repo jaromer/0bserverx-Qwen3.8-27B-Heretic-Abliterated-Uncensored-GGUF ./model-folderQwen/Qwen3.8-27B