← back to catalog · registered 2026-08-22 13:56

usermma/AMALIA-9B-0626-SFT-Heretic-Abliterated-mlx-fp16

usermma Llama 9.2B 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/usermma%2FAMALIA-9B-0626-SFT-Heretic-Abliterated-mlx-fp16"
Response includes
  • classification m3
  • files 12
  • hub_downloads_all_time 91
  • author_summary 77 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 · lifetime
91
12 last 30d - stable
Likes
0
Model age
3mo ago
created 2026-07-10
Downloads over time
Now93→from35↑166%
3254779935 on Jul 1593 on Oct 1193 on Oct 3JulAugSepOct
Jul 15 → Oct 11 · 53 snapshots · spans 88 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 · 22 downloads combined

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

Metadata

Tags
mlx safetensors llama obliteratus abliteration uncensored obliterate mlx-my-repo en base_model:usermma/AMALIA-9B-0626-SFT-Heretic-Abliterated base_model:finetune:usermma/AMALIA-9B-0626-SFT-Heretic-Abliterated region:us

Related

Total size
17.0 GB
Files
12
Quantizations
1
Registered
2026-08-22 13:56
Last updated on HF
2026-07-10 08:57

Files by quantization

Auxiliary files 12 files 17.1 GB
model-00002-of-00004.safetensors 4.98 GB 3be45e9a download
model-00001-of-00004.safetensors 4.98 GB 5f806826 download
model-00003-of-00004.safetensors 4.94 GB be8b11ad download
model-00004-of-00004.safetensors 2.15 GB 7c70c828 download
tokenizer.json 15.1 MB 5d9928d1 download
model.safetensors.index.json 32.2 KB 4d0fe34a download
.gitattributes 1.53 KB 52373fe2 download
README.md 1.05 KB eb6d8304 download
config.json 819 B be4cf772 download
tokenizer_config.json 634 B eb51a680 download
chat_template.jinja 624 B f3619310 download
generation_config.json 190 B 6728ea4b download

README current version from Hugging Face


language: en
tags:

  • obliteratus
  • abliteration
  • uncensored
  • obliterate
  • mlx
  • mlx-my-repo
    base_model: usermma/AMALIA-9B-0626-SFT-Heretic-Abliterated

usermma/AMALIA-9B-0626-SFT-Heretic-Abliterated-mlx-fp16

The Model usermma/AMALIA-9B-0626-SFT-Heretic-Abliterated-mlx-fp16 was converted to MLX format from usermma/AMALIA-9B-0626-SFT-Heretic-Abliterated using mlx-lm version 0.31.2.

Use with mlx

pip install mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("usermma/AMALIA-9B-0626-SFT-Heretic-Abliterated-mlx-fp16")

prompt="hello"

if hasattr(tokenizer, "apply_chat_template") and tokenizer.chat_template is not None:
    messages = [{"role": "user", "content": prompt}]
    prompt = tokenizer.apply_chat_template(
        messages, tokenize=False, add_generation_prompt=True
    )

response = generate(model, tokenizer, prompt=prompt, verbose=True)

README history 1 version

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

  1. 2026-07-10Upload folder using huggingface_hub30682dc1 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