license: apache-2.0
FP8 version of DavidAU's https://huggingface.co/DavidAU/Qwen3-4B-2507-Thinking-heretic-abliterated-uncensored
Quantized with llm-compressor.
isola-tropicale/Qwen3-4B-2507-Thinking-heretic-abliterated-uncensored-FP8
Your rig
guess connected
? Why do I need an app?
The browser can't see your hardware
For privacy reasons a browser tells us at most "≥ 8 GB RAM, 8 cores" - same reading whether you have 8 GB or 128 GB. It has no idea how much RAM is free right now, which apps are open, or whether you have a GPU.
The Abliteration app is integrated with your machine
It reads your exact RAM, GPU model and VRAM, free memory right now, and picks the sharpest quant that still fits. Every model page lights up precisely for your rig.
And you can chat with any model, right now
The app is a full local runtime - no API keys, no subscription, everything runs on your machine. Click any model on this site and start a conversation in seconds.
Get the free 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/isola-tropicale%2FQwen3-4B-2507-Thinking-heretic-abliterated-uncensored-FP8" Response includes
- classification m3
- files 13
- hub_downloads_all_time 82
- author_summary 1 models
- readme_text full
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
Confidence
HIGH
Why we say so
name contains 'heretic'; Heretic default extraction is difference-of-means (Arditi 2024)
Downloads · lifetime
82
19 last 30d
- stable
Likes
1
Model age
9mo ago
created 2025-12-19
Downloads over time
Dec 17, 2025 → Oct 11 · 82 snapshots · spans 298 days
Metadata
Files by quantization
Auxiliary files 13 files 4.13 GB
| model.safetensors | 4.11 GB | 9aa8b1b2 | download |
| tokenizer.json | 10.9 MB | 69208fd2 | download |
| vocab.json | 2.65 MB | 4783fe10 | download |
| merges.txt | 1.59 MB | 31349551 | download |
| tokenizer_config.json | 5.41 KB | e2f882cd | download |
| chat_template.jinja | 3.95 KB | 2e2f69c3 | download |
| config.json | 2.85 KB | 35bdd64c | download |
| .gitattributes | 1.53 KB | 52373fe2 | download |
| added_tokens.json | 707 B | b54f9135 | download |
| special_tokens_map.json | 613 B | ac23c0aa | download |
| generation_config.json | 214 B | 98e0755a | download |
| README.md | 169 B | 09789fa9 | download |
| recipe.yaml | 136 B | f87e7f4e | download |
README
README history
The author's README evolved over time. Click a version to see its content at that point.
2025-12-19Update README.md
Loading...2025-12-19initial commit
Loading...