← back to catalog · registered 2026-11-11 04:03

mlx-community/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-OptiQ-4bit

mlx-community Deepseek 9B 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/mlx-community%2FDeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-OptiQ-4bit"
Response includes
  • classification m3
  • files 15
  • 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
3K
Likes
6
Model age
1w ago
created 2026-11-11
Downloads over time
Now 3.3K → from 0 ↑ 0%
01.2K2.4K3.6K 0 on Nov 11 3.3K on Nov 22 Nov
Nov 11 → Nov 22 · 12 snapshots · spans 11 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.

Metadata

License
apache-2.0
Languages
en zh ko ja es ru
Tags
mlx safetensors qwen3_5 optiq quantized 4bit mixed-precision qwen3.5 reasoning distillation deepseek multimodal

Related

Total size
7.46 GB
Files
15
Quantizations
1
Registered
2026-11-11 04:03
Last updated on HF
2026-11-11 06:15

Files by quantization

Auxiliary files 15 files 7.48 GB
model-00001-of-00002.safetensors 4.98 GB 47ce6690 download
model-00002-of-00002.safetensors 1.63 GB 2591892f download
optiq_vision.safetensors 870 MB f79255e0 download
tokenizer.json 19.1 MB a5cd9732 download
model.safetensors.index.json 113 KB 20c90adb download
config.json 54.8 KB 942a9477 download
chat_template.jinja 8.53 KB 3318462a download
README.md 3.25 KB a6e90d4b download
.gitattributes 1.53 KB 52373fe2 download
tokenizer_config.json 1.53 KB 338ef614 download
processor_config.json 1.29 KB a9804dd9 download
video_preprocessor_config.json 817 B e32b1d90 download
preprocessor_config.json 781 B 484b50c4 download
kv_config.json 528 B 69ca6e63 download
generation_config.json 142 B 79b5436a download

README current version from Hugging Face


license: apache-2.0
base_model: xtools-at/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic
base_model_relation: quantized
library_name: mlx
pipeline_tag: image-text-to-text
language:

  • en
  • zh
  • ko
  • ja
  • es
  • ru
    tags:
  • mlx
  • optiq
  • quantized
  • 4bit
  • mixed-precision
  • qwen3.5
  • reasoning
  • distillation
  • deepseek
  • multimodal
  • vision
  • image-text-to-text
  • mtp
  • multi-token-prediction
  • function-calling
  • tool-use
  • heretic
  • uncensored
  • apple-silicon

mlx-community/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-OptiQ-4bit

Built with mlx-optiq, the MLX-native toolkit to quantize, fine-tune, and serve LLMs locally on Apple Silicon, no PyTorch and no cloud. All OptiQ quants · Docs

OptiQ mixed-precision quant of xtools-at/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic, an uncensored (heretic) version of a DeepSeek-V4-Pro reasoning distill into Qwen3.5-9B. It is a Qwen3.5-family vision-language model with a bundled MTP speculation head. 7.7 GB on disk.

What it is

Property Value
Base xtools-at/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic (Qwen3.5, 9B)
Method OptiQ mixed-precision, per-layer 4/8-bit
Bit allocation Reused from the Qwen3.5-9B OptiQ recipe: the architecture is identical, so the per-layer sensitivity ranking transfers directly and no per-model sweep is needed
Layer split 116 components at 4-bit, 134 at 8-bit
Group size 64
On disk 7.7 GB
MTP Speculation head preserved in optiq/mtp.safetensors
Vision bf16 vision tower kept in optiq_vision.safetensors for image input
KV cache Mixed-precision KV config in kv_config.json

Following the naming llama.cpp uses for its mixed quants, the "4bit" label denotes the family, not the weighted average.

Checked after conversion against the bf16 model on the same prompts (facts, arithmetic, code, explanation): the answers match.

Run it

pip install mlx-optiq
import optiq  # registers the arch + MTP/vision sidecars
from mlx_lm import load, generate

model, tok = load("mlx-community/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-OptiQ-4bit")
prompt = tok.apply_chat_template(
    [{"role": "user", "content": "Explain mixed-precision quantization in two sentences."}],
    tokenize=False, add_generation_prompt=True,
)
print(generate(model, tok, prompt=prompt, max_tokens=400))

For image input plus an OpenAI- and Anthropic-compatible endpoint with mixed-precision KV cache:

optiq serve --model mlx-community/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-OptiQ-4bit

This is a reasoning model, so give it a generous token budget.

LM Studio: its downloader flattens the optiq/ folder. Move mtp.safetensors back into optiq/ or the speculation head is not found.

Links

README history 2 versions

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

  1. 2026-11-11 Card: on-disk size in decimal GB, matching mlx-optiq.com 7e60c1f 3.3 KB
    Loading...
  2. 2026-11-11 Add files using upload-large-folder tool 957faa9 3.3 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