trohrbaugh / Model details

Qwen3.8-27B-heretic-ara

trohrbaugh/Qwen3.8-27B-heretic-ara

abliterateduncensored

Read the original model card ↗ · Choosing a model

About this model

trohrbaugh's decensored derivative of Qwen/Qwen3.8-27B, released as safetensors weights under Apache-2.0. The creator describes it as made with a custom fork of Heretic v1.2.0 using Arbitrary-Rank Ablation (ARA), applied to layers 26 through 56 with preserve_good_behavior_weight 0.9432, steer_bad_behavior_weight 0.0009, overcorrect_relative_weight 0.5038, and neighbor_count 10. The model card labels it abliterated/uncensored and reports a KL divergence of 0.0535 versus the original and 0 out of 100 refusals in its evaluation, while the original is listed at 99 out of 100. The repository is tagged as image-text-to-text and conversational, but claims about refusal reduction and retained behavior have not been independently verified.

Description sources

https://huggingface.co/trohrbaugh/Qwen3.8-27B-heretic-ara/raw/a67ae100d933c0d17af3232bda35825979fc63ce/README.md ↗

Source-based description, not an independent evaluation. Check the linked sources for the creator’s claims and limitations.

Size and formats

Base model: Unknown
Parameters: 27.3567B
Main repository formats: safetensors
License tag: apache-2.0
Languages: Unknown
Last updated: 2026-08-14 · 18:28:00.000 UTC

What changed

Model types are creator-reported, not independently tested. These labels describe source metadata, not a guarantee of behavior.

abliteratedsource tag: abliterated
uncensoredsource tag: uncensored

Family evidence

These groupings are source-reported. Explicit base identifiers take priority over generic architecture tags.

Qwen: Family / architecture tag: qwen3_5

Available versions

Versions appear here only when they explicitly name this exact model as their source. Check each repository for available files.

Download sources

ModelScope distribution ↗

Distribution checked: ModelScope · 2026-10-29 · 06:30:24 UTC

Distribution links are not independent reviews or runtime compatibility guarantees. Check each source’s exact files and instructions.

Try it online

Try on Featherless ↗

Listing checked: Featherless · 2026-10-29 · 06:30:24 UTC

Third-party hosted inference listing this exact model ID. Hosting settings, quantization and pricing are the host’s, not tested here.

Link to this listing

Made or quantized this model? Paste one of these into your model card or README so people can find this listing.

Markdown

[Listed on Custom Clankers](https://customclankers.com/models/trohrbaugh-qwen3-8-27b-heretic-ara/)

HTML

<a href="https://customclankers.com/models/trohrbaugh-qwen3-8-27b-heretic-ara/">Listed on Custom Clankers</a>

Plain text links only: no image, no script, nothing loaded from this site.

Customize or build on this model

Listed from public GitHub data, not tested by us. Read the code and license before you run anything.

Tools to customize it further

Remove more refusals

Heretic ↗ ★ 36,945
Removes refusal behavior from transformer models automatically. It pairs directional ablation (abliteration) with an Optuna parameter search that tries to cut refusals while keeping the model close to the original, so you don't have to pick layers by hand.
OBLITERATUS ↗ ★ 8,998
A larger refusal-removal toolkit with several abliteration methods, usable locally or through a hosted Hugging Face Space. Check the telemetry setting before you run it: the README describes telemetry as opt-in, but on by default in the Spaces version.

Fine-tune it on your own data

Unsloth ↗ ★ 78,677
A desktop app and Python library for running and training models locally. Its README lists LoRA, QLoRA, full fine-tuning, DPO and GRPO, plus export to formats including GGUF, on Windows, Linux, WSL and macOS.
LlamaFactory ↗ ★ 75,759
Fine-tunes more than 100 language and vision-language models from a command line or a browser-based Web UI, aimed at training without writing code.

Make a smaller download (quantize)

llama.cpp ↗ ★ 132,640
Runs models locally and includes the tools behind most GGUF files. A conversion script turns Hugging Face weights into GGUF, and llama-quantize shrinks a high-precision GGUF to lower-bit formats, which can cost some accuracy.
MLX LM ↗ ★ 7,495
Apple's MLX package for running and fine-tuning models on Apple silicon. It can quantize models and upload them to the Hugging Face Hub, and supports LoRA and full fine-tuning.

Merge it with another model

mergekit ↗ ★ 7,421
Merges pretrained models with many merge methods. It works out-of-core, so merges can run on a CPU or with as little as 8 GB of VRAM, and it also handles LoRA extraction and mixture-of-experts merges.

Learn the techniques

LLM Course ↗ ★ 84,739
A free course with roadmaps and Colab notebooks. Its notebooks include one-click tools for merging with MergeKit, quantizing to GGUF, GPTQ, EXL2, AWQ and HQQ, and automatic abliteration with custom datasets.

Compare all tools →

Source provenance

Downloads: 2,034
Likes: 113
Metadata checked: 2026-10-29 · 00:13:52 UTC
First seen in index: 2026-10-29 · 00:13:52 UTC
Source revision: a67ae100d933c0d17af3232bda35825979fc63ce

Source-card excerpt

Source model card excerpt: This is a decensored version of Qwen/Qwen3.8-27B, made using Heretic (custom fork) v1.2.0+custom with the Arbitrary-Rank Ablation (ARA) method Abliteration parameters Parameter Value :--------

From trohrbaugh/Qwen3.8-27B-heretic-ara. Creator model-card excerpt, not an independent review.

Inspect before you run

Low refusal is not proof of correctness. Outputs can be inaccurate, harmful, or biased. Review the model card, license and files yourself. A license tag is not commercial-use advice. Parameter counts are source metadata; no hardware compatibility is implied.

This directory does not host or redistribute model weights. Refusal profiles are measured by us on a fixed prompt set; all other behavior claims are source-reported.

Collection rules & limitations ↗