Get the app
← back to catalog · registered 2026-11-07 12:02

ZombieYU/Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-NINFER-V3

ZombieYU 27B GGUF 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/ZombieYU%2FQwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-NINFER-V3"
Response includes
  • classification m3
  • files 4
  • 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 · 30-day
0
Likes
0
Model age
today
created 2026-11-07

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.

DavidAU/Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-GGUF↓ 15,557 this lineage
ZombieYU/Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-NINFER-V3

Metadata

License
apache-2.0
Tags
ninfer image-text-to-text base_model:DavidAU/Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-GGUF base_model:finetune:DavidAU/Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-GGUF license:apache-2.0 region:us
Total size
0 B
Files
4
Quantizations
1
Registered
2026-11-07 12:02
Last updated on HF
2026-11-07 12:15
Source huggingface.co/ZombieYU/Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-NINFER-V3 ↗ Base model DavidAU/Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-GGUF →

Files by quantization

Auxiliary files 4 files 16.6 GB
Qwen3.8-27B-Ultimate-Heretic-Uncn-Turbo-Brilli-50x-AGI-NEO-MTP-Q4_K_M-Ninfer-v3.ninfer 16.6 GB 02bba170 download
Qwen3.8-27B-Ultimate-Heretic-Uncn-Turbo-Brilli-50x-AGI-NEO-MTP-Q4_K_M-converted.ninfer.conversion.json 5.47 KB b1ce44c3 download
README.md 1.98 KB 5de63fb8 download
.gitattributes 1.60 KB a152b247 download

README current version from Hugging Face

license: apache-2.0
base_model:

  • DavidAU/Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-GGUF

Qwen3.8-27B-Ultimate-Heretic-Uncensored (Ninfer v3 Format)

This is the Ninfer v3 converted version of the DavidAU/Qwen3.8-27B-Ultimate-Heretic-Uncensored-Turbo-Brilliance-50x-AGI-NEO-GGUF model.

I have just done the conversion (NInferEZ Converter) of Q4_K_M GGUF to Ninfer v3 with included vision model and MTP support.

🚀 Testing & Benchmark Setup

Tested with Ninfer-all on Windows 10 with full context size (262,144 tokens) on an RTX 3090.

Inference Command (Windows)

Run directly in CMD or PowerShell:

"Z:\ninfer-windows-x64-cuda13.4-sm75-80-86-89-120a-6d8dc3c1e0e1\ninfer-serve.exe" "Z:\Models\Qwen3.8-27B-Ultimate-Heretic-Uncn-Turbo-Brilli-50x-AGI-NEO-MTP-Q4_K_M-Ninfer-v3.ninfer" --host 0.0.0.0 --port 1234 --max-concurrency 3 --max-context 262144 --kv-capacity 262144 --kv-dtype rk4v4-e8 --spec mtp --draft-tokens 3 --embedding-q4 --gdn-state-fp16 --prefill-cublas --prefill-chunk 4096 --max-pending-requests 16 --pending-timeout-ms 600000 --vision --vision-residency overlay --max-private-continuations 8 --max-shared-prefixes 8 --host-state-slots 32 --host-kv-mib 8192 --auto-prefix-grid --model-id Qwen38Dense

⚠️ Thermal Note & Power Limitation (HP Z640)

I have received ~90 tps on the RTX 3090, but with restricted power of the GPU to 250W:

nvidia-smi -pl 250

This is necessary because the GPU Memory Junction Temperature goes over 100°C on an unmodified HP Z640 case with higher GPU power.

Recommendation: I advise you to check the temperature of the GPU memory with every model change (I use HWiNFO for checking and MSI Afterburner to increase fan speed).

🤝 Credits

All credits go to all people for their respective work!

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