← back to catalog · registered 2026-11-16 19:03

prithivMLmods/Qwen3.8-27B-Ultimate-Heretic-Uncensored-NVFP4

prithivMLmods 27B 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/prithivMLmods%2FQwen3.8-27B-Ultimate-Heretic-Uncensored-NVFP4"
Response includes
  • classification m3
  • files 15
  • author_summary 117 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
88
Likes
1
Model age
5d ago
created 2026-11-16

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.

Downloads over time
Now 88 → from 0 ↑ 0%
0326597 0 on Nov 16 88 on Nov 22 Nov
Nov 16 → Nov 22 · 7 snapshots · spans 6 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.

Variants by this author 2 formats · 1K downloads combined

The same weights this author released in different packaging. Pick the format that matches your runtime.

Metadata

License
apache-2.0
Languages
en
Tags
transformers safetensors qwen3_5 image-text-to-text mtp text-generation-inference unsloth tool-use heretic uncensored decensored abliterated

Related

Total size
19.1 GB
Files
15
Quantizations
1
Registered
2026-11-16 19:03
Last updated on HF
2026-11-16 18:27

Files by quantization

Auxiliary files 15 files 19.2 GB
model-00001-of-00004.safetensors 4.66 GB 5b45d5fa download
model-00003-of-00004.safetensors 4.63 GB 769ac736 download
model-00002-of-00004.safetensors 4.62 GB fc3c3609 download
model-00004-of-00004.safetensors 4.45 GB 6643c3c0 download
model_mtp.safetensors 810 MB 735f9bc2 download
tokenizer.json 19.1 MB a5cd9732 download
model.safetensors.index.json 272 KB f1d9127b download
chat_template.jinja 242 KB 30a3a5b3 download
config.json 10.1 KB 04131044 download
README.md 5.59 KB 36cfce27 download
.gitattributes 1.53 KB 52373fe2 download
processor_config.json 1.19 KB 43c4343e download
tokenizer_config.json 1.19 KB 86148f3d download
recipe.yaml 334 B a5b31245 download
generation_config.json 214 B a4b665dd download

README current version from Hugging Face


tags:

  • mtp
  • text-generation-inference
  • transformers
  • unsloth
  • qwen3_5
  • tool-use
  • heretic
  • uncensored
  • decensored
  • abliterated
  • speculative-decoding
  • nvidia-blackwell
  • rtx-pro-6000
  • SM120
  • GB202
  • B200
  • B300
  • FP4
  • blackwell
  • Qwen3.8
  • GPTQ
    datasets:
  • HuggingFaceH4/ultrachat_200k
    base_model:
  • DavidAU/Qwen3.8-27B-Ultimate-Heretic-Uncensored
    license: apache-2.0
    language:
  • en
    pipeline_tag: image-text-to-text
    library_name: transformers

Qwen3.8-27B-Ultimate-Heretic-Uncensored-NVFP4

Qwen3.8-27B-Ultimate-Heretic-Uncensored-NVFP4 is a high-efficiency NVFP4 quantization of DavidAU/Qwen3.8-27B-Ultimate-Heretic-Uncensored calibrated using GPTQ with an imatrix_mse weight observer (targeting Linear layers while preserving lm_head and visual modules in full precision). Built upon the multi-stage-tuned Qwen 3.8 27B architecture with DavidAU's Turbo Brilliance V2 50X AGI system and Heretic reasoning and instruct frameworks, this compressed variant integrates 27 reasoning modes, 27 instruct modes, and over 15 dynamic AI constructs to deliver low-footprint, high-throughput inference while maintaining deep refusal removal, multilingual capability, and token efficiency across complex creative, coding, and analytical workloads.

[!NOTE]
MTP — Speculative Decoding: Yes | Accelerated FP4 Tensor Core Performance

Quantization Recipe

The model is quantized using the GPTQModifier with the NVFP4 quantization scheme.

default_stage:
  default_modifiers:
    GPTQModifier:
      targets: [Linear]
      ignore: [lm_head, 're:.*visual.*']
      scheme: NVFP4
      weight_observer: imatrix_mse
      bypass_divisibility_checks: false
      requires_calibration_data: true
      block_size: 128
      dampening_frac: 0.01
      batched_quantization: auto

Model Files

File Name File Type File Size File Link Description
chat_template.jinja Jinja Template 248 kB Link Chat template for formatting conversations and prompts.
config.json JSON 10.3 kB Link Core model architecture and configuration parameters.
generation_config.json JSON 214 Bytes Link Default text generation parameters (e.g., temperature, top_p).
model-00001-of-00004.safetensors Safetensors 5 GB Link Model weight shard 1/4. Required to load the model.
model-00002-of-00004.safetensors Safetensors 4.96 GB Link Model weight shard 2/4. Required to load the model.
model-00003-of-00004.safetensors Safetensors 4.97 GB Link Model weight shard 3/4. Required to load the model.
model-00004-of-00004.safetensors Safetensors 4.78 GB Link Model weight shard 4/4. Required to load the model.
model.safetensors.index.json JSON 278 kB Link Index file mapping model weights to their respective shards.
model_mtp.safetensors Safetensors 849 MB Link Multi-Token Prediction (MTP) auxiliary model weights, used to speed up generation.
processor_config.json JSON 1.22 kB Link Configuration for the multimodal processor.
recipe.yaml YAML 334 Bytes Link Configuration file detailing the quantization or training recipe (e.g., NVFP4 parameters).
tokenizer.json JSON 20 MB Link Complete tokenizer vocabulary and encoding rules.
tokenizer_config.json JSON 1.22 kB Link Tokenizer-specific configuration parameters.

License and Attribution

This model is based on and/or incorporates the following open-source projects and models:

This model is released under the Apache License 2.0.

README history 8 versions

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

  1. 2026-11-16 Update README.md c2f0bbb 5.6 KB
    Loading...
  2. 2026-11-16 Update README.md f130cad 5.6 KB
    Loading...
  3. 2026-11-16 Update README.md 59fb854 5.2 KB
    Loading...
  4. 2026-11-16 Update README.md f8051d7 4.5 KB
    Loading...
  5. 2026-11-16 Update README.md 62d8899 975 B
    Loading...
  6. 2026-11-16 Update README.md 75b6c5a 987 B
    Loading...
  7. 2026-11-16 Update README.md 7b18029 886 B
    Loading...
  8. 2026-11-16 Create README.md 211f327 69 B
    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