← back to catalog · registered 2026-11-19 09:04

mlx-community/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-oQ4h-mtp

mlx-community Deepseek 9B
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/mlx-community%2FDeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-oQ4h-mtp"
Response includes
  • classification m3
  • files 16
  • author_summary 211 models
  • readme_text full
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
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

Confidence
HIGH
Why we say so
name contains 'heretic'; Heretic default extraction is difference-of-means (Arditi 2024)
Downloads · 30-day
213
Likes
0
Model age
3d ago
created 2026-11-19
Downloads over time
Now 213 → from 0 ↑ 0%
078156234 0 on Nov 19 213 on Nov 22 Nov
Nov 19 → Nov 22 · 4 snapshots · spans 3 days

Metadata

Tags
mlx safetensors qwen3_5 oq quantized mtp 4-bit region:us

Related

Total size
6.08 GB
Files
16
Quantizations
1
Registered
2026-11-19 09:04
Last updated on HF
2026-11-19 06:32

Files by quantization

Auxiliary files 16 files 6.10 GB
model-00002-of-00002.safetensors 3.74 GB 0505a8e4 download
model-00001-of-00002.safetensors 2.33 GB 6c5b30cf download
tokenizer.json 19.1 MB a5cd9732 download
vocab.json 4.99 MB b94df23d download
model.safetensors.index.json 145 KB 3ead133a download
config.json 37.0 KB 9f6b48b0 download
chat_template.jinja 8.53 KB 3318462a download
oq_imatrix_report.json 3.27 KB 62c93a36 download
.gitattributes 1.53 KB 52373fe2 download
processor_config.json 1.29 KB a9804dd9 download
tokenizer_config.json 1.19 KB 0bf8509e download
README.md 1.18 KB b01bec5d download
added_tokens.json 904 B 9955816d download
special_tokens_map.json 876 B fca0ded4 download
preprocessor_config.json 781 B 484b50c4 download
generation_config.json 142 B 79b5436a download

README current version from Hugging Face


library_name: mlx
tags:

  • mlx
  • oq
  • quantized
  • mtp

DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-oQ4h-mtp

[!NOTE]
Updated 2026-11-19: Re-quantized with oMLX v0.8.0. Please re-download if you have an earlier copy.

This model was quantized using oQ (oMLX v0.8.0) mixed-precision quantization.

The effective average is 13.759 bits/weight for the complete checkpoint.

Quantization and Bit Distribution

Storage format Logical weights Share of total Tensor storage Effective bits/weight
Affine 4-bit, group 64 1.156B 30.458% 0.605 GiB 4.50
Affine 5-bit, group 64 0.071B 1.879% 0.046 GiB 5.50
Affine 6-bit, group 64 1.017B 26.810% 0.770 GiB 6.50
BF16 0.600B 15.816% 1.118 GiB 16.00
U32 0.950B 25.037% 3.538 GiB 32.00
Total 3.794B 100% 6.077 GiB 13.759

Storage Breakdown

Component Logical weights Storage
Language model 3.095B 5.419 GB / 5.047 GiB
Vision 0.456B 0.912 GB / 0.849 GiB
MTP head 0.243B 0.194 GB / 0.181 GiB
Total 3.794B 6.525 GB / 6.077 GiB

README history 1 version

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

  1. 2026-11-19 Upload DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-oQ4h-mtp via oMLX 0d03c4f 1.2 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