← back to catalog · registered 2026-11-18 04:04

hhjuyh/Qwen3.8-35B-A3B-Hierophant-Abliterated

hhjuyh 35B GGUF MoE multimodal
Your rig guess connected
? Why do I need an app?
Reading your rig…

This is a rough estimate. Install the free app - we'll show exact numbers.

Reading real hardware from your app right now. Numbers below are exact.

Below is the per-quantization compatibility for this model.

curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/hhjuyh%2FQwen3.8-35B-A3B-Hierophant-Abliterated"
Response includes
  • classification m8
  • files 5
  • author_summary 3 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M8
Primary method

Repackaging (quantization)

Applied on top of direct removal inherited from the base model.
Confidence
MEDIUM
Inherited from base model
Why this label 3 signals
Method inferred from partial signals - repository name, related files, or tag patterns. Producer identity not confirmed; label may sharpen or shift as we gather more evidence.
  • 'abliterated' in name/tags
  • is_gguf=1
  • assume M1 (base ablation) + M8 (GGUF quant) - default when producer unknown
Refusal direction extraction

No specific extraction method could be identified for this model. The producer either did not document it or used a proprietary pipeline.

What is a refusal direction? →
Downloads · 30-day
551
↑ 69% in 90 days
Likes
0
Model age
1w ago
created 2026-11-14
Downloads over time
Now 551 → from 327 ↑ 69%
316402488573 327 on Nov 18 551 on Nov 22 Nov
Nov 18 → Nov 22 · 5 snapshots · spans 4 days

Genealogy 0 direct forks

Full fork graph →

This model's place in the market. Above: what it was derived from. Below: the tree of everything derived from it.

Metadata

License
apache-2.0
Languages
en
Quantizations
Q4_K Q6_K
Tags
llama.cpp gguf qwen3.8 hierophant abliteration quantized image-text-to-text en base_model:empero-ai/Qwen3.8-35B-A3B-Distill-GGUF base_model:quantized:empero-ai/Qwen3.8-35B-A3B-Distill-GGUF license:apache-2.0 endpoints_compatible

Related

Total size
47.4 GB
Files
5
Quantizations
3
Registered
2026-11-18 04:04
Last updated on HF
2026-11-15 14:43

Files by quantization

Q6_K 1 file 27.2 GB
Qwen3.8-35B-A3B-Hierophant-1.0-Q6_K.gguf 27.2 GB bb883d03 download
Q4_K 1 file 20.2 GB
Qwen3.8-35B-A3B-Hierophant-1.0-Q4_K_M.gguf 20.2 GB 2d600c1f download
Auxiliary files 3 files 15.4 KB
LICENSE 11.1 KB 1d5180a4 download
README.md 2.69 KB 89329143 download
.gitattributes 1.64 KB 89dabbf7 download

README current version from Hugging Face


license: apache-2.0
base_model: empero-ai/Qwen3.8-35B-A3B-Distill-GGUF
library_name: llama.cpp
pipeline_tag: image-text-to-text
language:

  • en
    tags:
  • gguf
  • qwen3.8
  • hierophant
  • abliteration
  • quantized

Qwen3.8-35B-A3B-Distill Hierophant 1.0 GGUF

Upstream model: Empero Qwen3.8-35B-A3B-Distill-GGUF

This unofficial variant merges a rank-1 Hierophant adapter into Empero's original Q6_K GGUF at scale 1.0 to reduce refusal behavior. The merge edits 121 tensors and preserves the original formats for all 753 tensors. It is a standalone GGUF. Original Q6_K · Hierophant

Model file Quantization Size
Qwen3.8-35B-A3B-Hierophant-1.0-Q6_K.gguf Q6_K 29.21 GB
Qwen3.8-35B-A3B-Hierophant-1.0-Q4_K_M.gguf Q4_K_M 21.71 GB

Text and vision

Load the standalone GGUF in llama.cpp or import it into LM Studio. For image input, place the original mmproj-Qwen3.8-35B-A3B-F16.gguf alongside the model. A llama.cpp smoke test passed a text arithmetic prompt and read a title from an image using this projector. Download the upstream vision projector

Refusals and quality

The original and merged Q6_K builds each ran the same 36 prompts in llama.cpp with a 2,048-token context, thinking disabled, and temperature set to 0.

Result Original Q6_K Hierophant 1.0 Q6_K
Explicit refusals on deception and cheating prompts 10/10 0/10
Requests redirected to an honest alternative 0/10 1/10
Benign control prompts answered 4/4 4/4
Short answers correct 14/16 14/16
Multi-step JSON tasks correct 0/6 0/6
Total quality tasks correct 14/22 14/22

The merge began fulfilling nine of the ten deception and cheating requests. On the plagiarism prompt, it drafted an admission. Nine merged replies in that group reached the 192-token cap. The scores describe this small test set with thinking disabled. Long-context and tool-use performance were outside its scope, and vision received a single smoke test.

License and attribution

The weights are distributed under the Apache License 2.0. Empero released the distill and upstream GGUF builds, based on Qwen's Qwen3.6-35B-A3B. The export here applies the Hierophant weight modification described above. Empero distill · Qwen base model

README history 3 versions

The author's README evolved over time. Click a version to see its content at that point.

  1. 2026-11-15 List Q4_K_M in the model files table 63a656b 2.7 KB
    Loading...
  2. 2026-11-15 Link to Empero's upstream Qwen 3.8 GGUF repository 67599ab 2.6 KB
    Loading...
  3. 2026-11-14 Add Qwen 3.8 model card and upstream license 85a96a5 2.5 KB
    Loading...
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration