Get the app
← back to catalog · registered 2026-09-07 20:55

Madras1/Qwen3.8-27B-Abliterated-v2

Madras1 Qwen 27B
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/Madras1%2FQwen3.8-27B-Abliterated-v2"
Response includes
  • classification m1
  • files 36
  • author_summary 5 models
  • readme_text full
Try next
GET/models?author=Madras1
All 5 models by Madras1
 GET/models?family=qwen&method=m1
Sibling qwen models · same method
 GET/methods/m1
What "m1" means · taxonomy
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M1
Primary method

Direct removal

No other method signals detected in this model.
Confidence
MEDIUM
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=0 (base model)
  • no specific method indicators - defaulting to M1 (most common)
How we label → About M1 →
Refusal direction extracted via
Extraction technique

Difference-of-means

FromArditi et al. (2024) ↗
Confidence
MEDIUM
Why we say so
primary_method=M1; difference-of-means is the reference extraction for M1/M3 (Arditi 2024)
What is a refusal direction? → Read the paper ↗
Downloads · 30-day
0
Likes
0
Model age
6w ago
created 2026-09-07
Downloads over time
Now0→from0↑0%
00110 on Sep 70 on Sep 8Sep
Sep 7 → Sep 8 · 2 snapshots · spans 1 day

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.

Qwen/Qwen3.8-27B this lineage
Madras1/Qwen3.8-27B-Abliterated-v2

Metadata

License
apache-2.0
Tags
transformers safetensors qwen3_5_text text-generation qwen abliterated representation-engineering conversational base_model:Qwen/Qwen3.8-27B base_model:finetune:Qwen/Qwen3.8-27B license:apache-2.0 endpoints_compatible

Related

Other by Madras1
Qwen3.8-27B-Abliterated-Uncensored 491 Qwen3.8-27B-Abliterated 126 Gemma-4-E4B-Abliterated-Uncensored 114 Gemma-4-31B-Abliterated-Uncensored 13
Same base model
JonathanColetti/Qwen3.8-27B-Uncensored-GGUF 2.5M huihui-ai/Huihui-Qwen3.8-27B-abliterated-GGUF 2.4M HauhauCS/Qwen3.8-27B-Uncensored-HauhauCS-Aggressive-MTP-GGUF 2.3M 0bserverx/Qwen3.8-27B-Heretic-Abliterated-Uncensored-GGUF 1M azukivc/Qwen3.8-27B-Uncensored-GGUF 133K
Total size
50.1 GB
Files
36
Quantizations
1
Registered
2026-09-07 20:55
Last updated on HF
2026-09-07 20:57
Source huggingface.co/Madras1/Qwen3.8-27B-Abliterated-v2 ↗ Base model Qwen/Qwen3.8-27B ↗

Files by quantization

Auxiliary files 36 files 50.2 GB
model-00003-of-00006.safetensors 9.25 GB d96c3113 download
model-00004-of-00006.safetensors 9.22 GB a9ca7e2e download
model-00005-of-00006.safetensors 9.22 GB d692518e download
model-00001-of-00006.safetensors 9.22 GB 1d5a6418 download
model-00002-of-00006.safetensors 9.17 GB 9bb0c7a0 download
model-00006-of-00006.safetensors 4.03 GB a5d72d0a download
candidate_directions.pt 33.8 MB 816b07d2 download
tokenizer.json 19.1 MB 4b57ae28 download
evaluation_runtime_comparison.json 236 KB 8813d5c3 download
evaluation_comparison.json 235 KB 62f80473 download
abliteration_report.json 196 KB a69b003f download
direction_selection.json 174 KB 246a98d1 download
calibration_baseline.jsonl 119 KB 068952d6 download
direction_selection_progress.jsonl 106 KB eb622b76 download
model.safetensors.index.json 81.9 KB 44343183 download
evaluation_before.jsonl 73.9 KB f9df2846 download
evaluation_runtime_ablation.jsonl 70.4 KB 4faf9db4 download
evaluation_after.jsonl 70.2 KB 498c9a62 download
split_manifest.json 53.1 KB 32d13a7e download
weight_edit_audit.json 33.9 KB 124a86ff download
direction_geometry_progress.jsonl 16.2 KB 6c222ca1 download
chat_template.jinja 8.74 KB c0c686f9 download
baseline_filter_report.json 8.67 KB 7ff76e38 download
config.json 2.68 KB dddd9d63 download
.gitattributes 1.53 KB 52373fe2 download
runtime_quality_gate.json 1.32 KB 029ef57c download
baked_quality_gate.json 1.30 KB d67e1bbc download
tokenizer_config.json 1.10 KB ed1f99f3 download
experiment_config.json 979 B 28522268 download
evaluation_before_summary.json 709 B 0342695b download
README.md 568 B 4d270d1c download
refusal_token_calibration.json 237 B 6d13f974 download
generation_config.json 214 B 1f489fac download
evaluation_input_audit.json 122 B 1dc929d9 download
baseline_completion_loss.json 113 B ddd751e8 download
run_status.json 26.0 B 89715ebd download

README current version from Hugging Face

license: apache-2.0
base_model: Qwen/Qwen3.8-27B
library_name: transformers
pipeline_tag: text-generation
tags:

  • qwen
  • abliterated
  • representation-engineering

Qwen3.8-27B Abliterated v2

Checkpoint BF16 de texto derivado de Qwen/Qwen3.8-27B por intervenção
de representação.

A execução selecionou resid_pre na posição -5, camada 33, editou 129
módulos e passou os gates de capacidade e integridade.

A intervenção reduziu fortemente as recusas, mas não garante ausência
absoluta de alinhamento residual. Este é um checkpoint experimental.

README history 2 versions

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

  1. 2026-09-13Rename README_QWEN38.md to README.mdd45bdd49 KB
    Loading...
  2. 2026-09-07Upload Qwen3.8-27B Abliterated v2a8343ed568 B
    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.

Get the app Browse the catalog
Open in Abliteration