← back to catalog · registered 2026-11-13 18:03

esatapedico/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-NVFP4

esatapedico 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/esatapedico%2FDeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-NVFP4"
Response includes
  • classification m3
  • files 12
  • author_summary 16 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
80
Likes
0
Model age
1w ago
created 2026-11-13
Downloads over time
Now 80 → from 0 ↑ 0%
0295988 0 on Nov 13 80 on Nov 22 80 on Nov 21 Nov
Nov 13 → Nov 22 · 10 snapshots · spans 9 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
Tags
transformers safetensors qwen3_5 image-text-to-text nvfp4 deepseek vision multimodal blackwell compressed-tensors llama.cpp text-generation

Related

Total size
10.9 GB
Files
12
Quantizations
1
Registered
2026-11-13 18:03
Last updated on HF
2026-11-13 17:10

Files by quantization

Auxiliary files 12 files 10.9 GB
model.safetensors 10.4 GB ca824666 download
model_mtp.safetensors 464 MB e9f83dae download
tokenizer.json 19.1 MB a5cd9732 download
model.safetensors.index.json 80.7 KB 3ba6c2f2 download
config.json 15.9 KB 4383d9a9 download
chat_template.jinja 8.53 KB 3318462a download
README.md 5.56 KB 7cc8403b download
.gitattributes 1.53 KB 52373fe2 download
tokenizer_config.json 1.50 KB eeb27c42 download
processor_config.json 1.29 KB a9804dd9 download
recipe.yaml 251 B 5daf2c33 download
generation_config.json 147 B 3057bb05 download

README current version from Hugging Face


license: apache-2.0
base_model: xtools-at/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic
pipeline_tag: text-generation
library_name: transformers
description: NVFP4 (W4A16, compressed-tensors) self-conversion of DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic with MTP head and vision tower intact.
tags:

  • nvfp4
  • qwen3_5
  • deepseek
  • vision
  • multimodal
  • blackwell
  • compressed-tensors
  • llama.cpp
    language:
  • en
    quantization_config:
    quant_method: compressed-tensors
    format: nvfp4-pack-quantized

DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-NVFP4

NVFP4 checkpoint of xtools-at/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic, a DeepSeek-V4-Pro distill of Qwen3.5-9B (32-block hybrid: 24 linear-attention + 8 full-attention blocks).

HF repo: esatapedico/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-NVFP4

NVFP4 here means W4A16 with FP8 E4M3 per-tensor scales, group size 16, weight only, datafree (no calibration corpus). The MLP feed-forward weights and the full-attention q/k/v/output projections are quantized; everything else stays at full precision.

The llama.cpp-ready GGUF family (seven tiers), built from this checkpoint, lives in esatapedico/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-NVFP4-MTP-GGUF with its own card.

At a glance

Field Value
Format compressed-tensors nvfp4-pack-quantized
Quantization W4A16 ("NVFP4A16"), group size 16, FP8 E4M3 scales, weight only
Calibration none (datafree)
Quantized 96 MLP projections (24 linear-attn blocks) + 32 full-attention q/k/v/o (8 full-attn blocks) = 128 packed NVFP4 tensors
Kept in BF16 linear-attention SSM path (attn_gate, attn_qkv, ssm_alpha, ssm_beta, ssm_out, ssm_conv), all norms, token embeddings, lm_head, vision tower (333 tensors)
MTP head 15 mtp.* tensors carried through in model_mtp.safetensors, bit-verified against the source file
Base xtools-at/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic (9B hybrid VLM, Apache-2.0)

Quantized scope in detail

Per linear-attention block (24 blocks): mlp.gate_proj, mlp.up_proj, mlp.down_proj (4 packed groups each: gate/up/down + 1 scale companion pattern per the classic recipe). Per full-attention block (8 blocks at interval 4): self_attn.q_proj, k_proj, v_proj, o_proj. After quantization the packed census is exactly 128 NVFP4 tensors, and every unquantized tensor is byte-identical to the source. The 15 MTP tensors bypass the compressor (known llm-compressor MTP-drop behavior) and are restored from the source's standalone MTP file with bitwise verification.

Checkpoint

  • Two safetensors files: model.safetensors (11,198,705,136 bytes, SHA-256 ca824666fdbd23a01bf7d41659a450f2b03180cfc41e75ce6f98155b5559915b) + model_mtp.safetensors (486,582,848 bytes, SHA-256 e9f83dae0fed4d0633fba652cd7eb942642491a283b11072c3cfc77731e0fae9), under model.safetensors.index.json (1,031 mapped names: 775 trunk + 333 vision + 15 MTP + lm_head + misc — vision and MTP included, nothing dropped).
  • config.json with quantization_config (quant_method: compressed-tensors, format: nvfp4-pack-quantized), arch Qwen3_5ForConditionalGeneration (Qwen3.5 hybrid text backbone + vision tower).
  • Tokenizer, chat template (chat_template.jinja), preprocessor/processor configs intact.
  • recipe.yaml: the exact llm-compressor recipe used (datafree, classic patterns + ignores).

Serving

  • llama.cpp: use the sibling GGUF repo. A compressed-tensors checkpoint is not directly loadable by llama.cpp; the verified end-to-end path is BF16 → this checkpoint → GGUF (with convert_hf_to_gguf.py --outtype auto, NVFP4 kept native, GGML type 40).
  • vLLM / compressed-tensors engines: the distribution format mirrors published compressed-tensors NVFP4 checkpoints. Load should be auto-detected from the config. We have not exercised a vLLM load ourselves. Our verification was the byte-identity guard, the GGUF path, and full quality/speed benches on the GGUF family; treat a fresh vLLM load as untested until tried.

Verified with these tiers (GGUF form): PPL 4.42–4.53, GSM8K 0.875–0.905 (n=200), MMLU 63.9–68.3% (n=3000), IFEval 30/30 parsed, ToolCap 100% with zero fabricated tools, on two 16 GB Blackwell cards. Details and speed measurements in the GGUF family card.

Provenance

Derivative work, Apache-2.0 throughout; nothing here was trained or fine-tuned. Credit:

  1. xtools-at for the Heretic decensor tune, and Jackrong for the DeepSeek-V4-Pro-Qwen3.5-9B base.
  2. Red Hat AI / vLLM project for llm-compressor, used for the NVFP4 quantization.
  3. esatapedico for the conversion, verification (byte-identity guard, preflight, packed census), and benchmarking.

Follow along & support

I post updates on new conversions, benchmarks, and what I'm working on over on my site. Follow along there to keep up with new releases and the work in progress. If you'd like to support more of it, you will find the options there. I do this on consumer hardware and like seeing how far it goes. More is on the way.

🌐 esatapedico.com. Updates, work-in-progress, and ways to support.

Note on this card

This model card was written by an AI assistant at the request of the repository author, who did the engineering. As with any AI-generated text, there may be errors; please verify anything important (sizes, configs, commands) against the checkpoint files before relying on it.

README history 1 version

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

  1. 2026-11-13 Upload README.md with huggingface_hub 8e9b606 5.6 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