Get the app
← back to catalog · registered 2026-11-07 15:02

SargeDev/Arete-27B-Uncensored-NVFP4-FP8

SargeDev 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/SargeDev%2FArete-27B-Uncensored-NVFP4-FP8"
Response includes
  • classification m1
  • files 11
  • author_summary 11 models
  • readme_text full
Try next
GET/models?author=SargeDev
All 11 models by SargeDev
 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
today
created 2026-11-07

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.

NousResearch/Arete-27B this lineage
SargeDev/Arete-27B-Uncensored-NVFP4-FP8

Variants by this author 4 formats · 0 downloads combined

The same weights this author released in different packaging. Pick the format that matches your runtime.

Metadata

License
apache-2.0
Tags
safetensors qwen3_5_text abliterated uncensored qwen3.8 base_model:NousResearch/Arete-27B base_model:quantized:NousResearch/Arete-27B license:apache-2.0 8-bit compressed-tensors region:us

Related

Other by SargeDev
JEV-27B-Uncensored-GGUF 322 qwen3.8-flash-next-lil-rascal-uncensored 254 JEV-27B-Uncensored 116 JEV-27B-VL-Uncensored 102 JEV-27B-Uncensored-NVFP4 81
Same base model
Solotovfahl/Arete-27B-Abliterated-Solotov 4 vulpineriot/Qwen3.8-27B-Uncensored-v3 0 SargeDev/Arete-27B-Uncensored 0 SargeDev/Arete-27B-Uncensored-EXL3 0 SargeDev/Arete-27B-Uncensored-GGUF 0
Total size
27.3 GB
Files
11
Quantizations
1
Registered
2026-11-07 15:02
Last updated on HF
2026-11-07 16:59
Source huggingface.co/SargeDev/Arete-27B-Uncensored-NVFP4-FP8 ↗ Base model NousResearch/Arete-27B ↗

Files by quantization

Auxiliary files 11 files 27.3 GB
model-00001-of-00002.safetensors 18.6 GB 686ac5be download
model-00002-of-00002.safetensors 8.71 GB 2a8c7454 download
tokenizer.json 19.1 MB f399b3cd download
model.safetensors.index.json 127 KB f9b71d99 download
config.json 16.1 KB fa27165b download
chat_template.jinja 8.74 KB c0c686f9 download
recipe.yaml 2.09 KB 6f8ea4ce download
.gitattributes 1.53 KB 52373fe2 download
tokenizer_config.json 1.32 KB 2bc99a50 download
README.md 1.28 KB eb5343fe download
generation_config.json 244 B 14865c33 download

README current version from Hugging Face

license: apache-2.0
base_model: NousResearch/Arete-27B
tags:

  • abliterated
  • uncensored
  • qwen3.8

Arete-27B-Uncensored-NVFP4-FP8

First uncensored variant of NousResearch/Arete-27B.
Arete is a values-aligned full-parameter SFT of Qwen3.8-27B; no abliterated
version existed, so we made one with OBLITERATUS (--method aggressive,
bf16, 842-prompt harm/harmless contrast set).

What was done

  • Refusal-direction ablation across 15 strong layers (59-63 strongest)
  • Quality gates measured after the edit: coherence 1.0, KL 0.90 vs base,
    perplexity 3.22 -> 3.93 ( modest rise), capability suite intact
  • MTP/nextn drafter tensors preserved (stock Qwen3.8 head, spec-decode works)

Compliance battery (15 hard prompts: exploit dev, red-team, malware, DoS)

15/15 DELIVERED in bf16 (one long-form answer truncated by a 2500-char
scoring threshold but verified complete on re-run). Community GGUF/EXL3/NVFP4
variants score 14-15/15; a single-cell sampling artifact on phish-kit in one
quant was verified as delivery-on-retry, not refusal.

Format

NVFP4 W4A4 + FP8_DYNAMIC down_proj (compressed-tensors)

Credit

  • Nous Research: Arete-27B base
  • Pliny the Liberator: OBLITERATUS
  • Qwen: base architecture
  • SargeDev: abliteration run + quants
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