Get the app
← back to catalog · registered 2026-09-11 18:55

xy10086/Qwen3.8-9B-Distill-Uncensored-Heretic-MTPLX-Q6G64

xy10086 9.0B multimodal second-order
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/xy10086%2FQwen3.8-9B-Distill-Uncensored-Heretic-MTPLX-Q6G64"
Response includes
  • classification m3
  • files 18
  • hub_downloads_all_time 1,397
  • author_summary 1 models
  • readme_text full
Try next
GET/methods/m3
What "m3" means · taxonomy
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M3
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)
How we label → About M3 →
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)
What is a refusal direction? → Read the paper ↗
Downloads · lifetime
1K
HF trending score
1
mild recent interest
Likes
3
Model age
6w ago
created 2026-09-11
Downloads over time
Now3.1K→from0↑0%
01.1K2.3K3.4K0 on Sep 113.1K on Oct 27SepOct
Sep 11 → Oct 27 · 47 snapshots · spans 46 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.

petruhonk/Qwen3.8-9B-Distill-uncensored-heretic↓ 1,806 this lineage
xy10086/Qwen3.8-9B-Distill-Uncensored-Heretic-MTPLX-Q6G64↓ 1,869

Metadata

License
apache-2.0
Languages
en zh
Tags
mlx safetensors qwen3_5 qwen qwen3.8 mtplx apple-silicon mtp speculative-decoding multimodal vision quantized

Related

Same base model
ge525/Qwen3.8-9B-Distill-uncensored-heretic-Q4_K_M-GGUF 17K nuofang/Qwen3.8-9B-Distill-uncensored-heretic-GGUF 2K kataguru/Qwen3.8-9B-Distill-uncensored-heretic-Finnish-GGUF 981 Atomic-Germ/Qwen3.8-9B-Distill-uncensored-heretic-Q4_K_M-GGUF 461 985love211/Qwen3.8-9B-Distill-uncensored-heretic-Q5_0-GGUF 277
Total size
8.08 GB
Files
18
Quantizations
1
Registered
2026-09-11 18:55
Last updated on HF
2026-09-11 18:16
Source huggingface.co/xy10086/Qwen3.8-9B-Distill-Uncensored-Heretic-MTPLX-Q6G64 ↗ Base model petruhonk/Qwen3.8-9B-Distill-uncensored-heretic →

Files by quantization

Auxiliary files 18 files 8.10 GB
model-00001-of-00002.safetensors 4.99 GB 07b0dda7 download
model-00002-of-00002.safetensors 1.79 GB f9b62b84 download
model-vision.safetensors 870 MB 3db643f3 download
mtp.safetensors 464 MB 0919fab4 download
tokenizer.json 19.1 MB 06b95093 download
model.safetensors.index.json 113 KB d27d8ffd download
chat_template.jinja 7.57 KB a585dec8 download
mtplx_runtime.json 3.93 KB 5cc217cc download
config.json 3.56 KB ba29714e download
README.md 3.14 KB e8a38767 download
.gitattributes 1.53 KB 52373fe2 download
tokenizer_config.json 1.17 KB ffb89ca0 download
processor_config.json 1.16 KB 33818c7f download
benchmarks.json 810 B f18ae7c2 download
preprocessor_config.json 390 B 2ea84a43 download
video_preprocessor_config.json 385 B 3ba673a5 download
generation_config.json 164 B aaaf57bd download
forge-recipe.json 100 B 0da39ebc download

README current version from Hugging Face

license: apache-2.0
base_model: petruhonk/Qwen3.8-9B-Distill-uncensored-heretic
pipeline_tag: image-text-to-text
language:

  • en
  • zh
    tags:
  • qwen
  • qwen3.8
  • mtplx
  • mlx
  • apple-silicon
  • mtp
  • speculative-decoding
  • multimodal
  • vision
  • quantized
  • 6-bit
  • uncensored

Qwen3.8-9B-Distill-Uncensored-Heretic-MTPLX-Q6G64

MTPLX / Apple Silicon optimized Q6G64 conversion of:

petruhonk/Qwen3.8-9B-Distill-uncensored-heretic

Repository:

xy10086/Qwen3.8-9B-Distill-Uncensored-Heretic-MTPLX-Q6G64

Highlights

  • Qwen3.8 9B multimodal architecture
  • Q6 affine quantization, group size 64
  • Native BF16 MTP head preserved
  • 15/15 MTP tensors
  • BF16 vision tower preserved
  • 333 vision tensors
  • Image input verified
  • MTP speculative decoding verified
  • Recommended MTP depth: 2
  • Tested on Apple M5 Pro, 48 GB unified memory

Provenance

Source model:

petruhonk/Qwen3.8-9B-Distill-uncensored-heretic

Upstream model:

empero-ai/Qwen3.8-9B-Distill

The source repository did not include the vision preprocessing sidecars
required by MTPLX. Matching preprocessor_config.json and
video_preprocessor_config.json were restored from the upstream
empero-ai/Qwen3.8-9B-Distill repository.

The source vision weights were preserved by MTPLX Forge and stored in
model-vision.safetensors.

Quantization

Main language-model trunk:

  • 6-bit
  • Group size 64
  • Affine quantization

MTP:

  • BF16
  • 15 tensors
  • Not requantized

Vision:

  • BF16
  • 333 tensors
  • About 870 MiB on disk

Forge recipe: forge-recipe.json

Tested software

  • MTPLX 2.11.2
  • MLX 0.32.2
  • mlx-lm 0.31.3

Recommended configuration

  • Profile: sustained
  • MTP depth: 2
  • Context window: 32768
  • Max response tokens: 8192

Benchmarks

MTPLX Forge verification

Mode Decode tok/s
AR 32.42
D1 58.91
D2 71.70
D3 70.90

Real Chinese chat

Depth Avg decode tok/s Avg TTFT
D1 49.41 0.159 s
D2 50.73 0.134 s
D3 46.38 0.135 s

Vision test

  • 869 prompt tokens
  • 336 completion tokens
  • 65.30 decode tok/s
  • Successful screenshot understanding and OCR
  • Approx. 10.53 GB peak runtime memory

See benchmarks.json for the machine-readable summary.

Usage

Pull:

mtplx pull xy10086/Qwen3.8-9B-Distill-Uncensored-Heretic-MTPLX-Q6G64

Serve:

mtplx serve \
  --model xy10086/Qwen3.8-9B-Distill-Uncensored-Heretic-MTPLX-Q6G64 \
  --host 127.0.0.1 \
  --port 8000 \
  --profile sustained \
  --generation-mode mtp \
  --depth 2 \
  --context-window 32768 \
  --max-tokens 8192 \
  --reasoning off \
  --reasoning-effort low \
  --no-auth \
  --yes

OpenAI-compatible endpoint:

http://127.0.0.1:8000/v1

Behavior

This model is derived from a low-refusal / censorship-ablated checkpoint.
Lower refusal behavior does not imply higher factual accuracy.

Medical, legal, financial and other high-stakes outputs should be
independently verified.

Credits

  • petruhonk/Qwen3.8-9B-Distill-uncensored-heretic
  • empero-ai/Qwen3.8-9B-Distill
  • MTPLX

License

Apache-2.0. Review the upstream repositories for their attribution and
license requirements as well.

README history 1 version

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

  1. 2026-09-11Initial MTPLX Q6G64 release8aab4ee3.1 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.

Get the app Browse the catalog
Open in Abliteration