← back to catalog · registered 2026-08-22 13:56

amit21186/Qwen3.6-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking-6bit-MLX

Name: amit21186/Qwen3.6-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking-6bit-MLX
Author: amit21186

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/amit21186%2FQwen3.6-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking-6bit-MLX"

Response includes

classification m3
files 14
hub_downloads_all_time 1,290
author_summary 1 models
readme_text full

10 credits · hourly refresh · ~4 KB payload Get an API key →

Abliteration classifier · v1.0.0

Primary method

Layer-wise ablation

Applied on top of direct removal inherited from the base model.

Confidence

HIGH

Inherited from base model

M1 Direct removal →

Why this label 2 signals

Producer identity confirmed by naming conventions, tags or the model card. This label is very unlikely to change.

'heretic' in model name (Heretic-produced)
Heretic uses layer-wise optimization (M3) with underlying direction removal (M1)

Refusal direction extracted via

Extraction technique

Difference-of-means

FromArditi et al. (2024) ↗

Confidence

HIGH

Why we say so

name contains 'heretic'; Heretic default extraction is difference-of-means (Arditi 2024)

Downloads · lifetime

176 last 30d - stable

Likes

Model age

2mo ago

created 2026-08-05

Training datasets

Corpora the author lists in the model card. Datasets tracked in our /datasets catalog carry a category badge linking to the workflow stage. Others open on Hugging Face.

TeichAI/claude-4.5-opus-high-reasoning-250x DavidAU/PkDick-Deckard-5-Datasets

Downloads over time

Aug 5 → Oct 11 · 50 snapshots · spans 67 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.

DavidAU/Qwen3.6-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking↓ 536 this lineage

amit21186/Qwen3.6-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking-6bit-MLX↓ 176

Metadata

License

apache-2.0

Languages

en zh

Files by quantization

Auxiliary files 14 files 29.6 GB

model-00001-of-00006.safetensors	5.00 GB	1068de09	download
model-00002-of-00006.safetensors	4.98 GB	ac079e8a	download
model-00005-of-00006.safetensors	4.97 GB	55439329	download
model-00003-of-00006.safetensors	4.97 GB	193a1012	download
model-00004-of-00006.safetensors	4.97 GB	6a8e0d10	download
model-00006-of-00006.safetensors	4.70 GB	75d44b62	download
tokenizer.json	19.1 MB	06b95093	download
model.safetensors.index.json	278 KB	614bd341	download
chat_template.jinja	7.58 KB	a8755d82	download
config.json	5.08 KB	7a6e1af3	download
README.md	1.79 KB	341fc395	download
.gitattributes	1.53 KB	52373fe2	download
tokenizer_config.json	1.13 KB	8d610519	download
generation_config.json	202 B	023756cf	download

README current version from Hugging Face

language:

en
zh
license: apache-2.0
tags:
mlx
6-bit
quantized
uncensored
heretic
roleplaying
creative-writing
storytelling
fiction
datasets:
TeichAI/claude-4.5-opus-high-reasoning-250x
DavidAU/PkDick-Deckard-5-Datasets
pipeline_tag: text-generation
library_name: mlx
base_model: DavidAU/Qwen3.6-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking

Qwen3.6-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking — 6-bit MLX

This is a 6-bit MLX quantization of DavidAU/Qwen3.6-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking, converted with Apple's mlx-lm for fast local inference on Apple Silicon.

Credit

All credit for the underlying model, fine-tuning, and merge work goes to DavidAU. This repository only contains a quantized MLX conversion. See the original model card for full details on training data, intended use, and capabilities.

Released under the same license as the base model: Apache 2.0.

Quantization details

Method: mlx_lm.convert with -q --q-bits 6 (affine quantization, group size 64)
Actual bits per weight after quantization: 6.501 (group/scale overhead)
Source dtype: bfloat16
Converted on an Apple M1 Max, 64 GB unified memory

Use with mlx-lm

pip install mlx-lm
mlx_lm.generate --model amit21186/Qwen3.6-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking-6bit-MLX --prompt "Your prompt here"

Use with LM Studio

Search this repo name in LM Studio's model search, or download and place the folder under LM Studio's local models directory.

README history 1 version

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

2026-08-05Add 6-bit MLX quantization of Qwen3.6-40B-Claude-4.6-Opus-Deckard-Heretic-Unc...b00c8481.8 KB
Loading...