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

Babsie/Mag-Mell-R1-Uncensored-21B

Name: Babsie/Mag-Mell-R1-Uncensored-21B
Author: Babsie

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/Babsie%2FMag-Mell-R1-Uncensored-21B"

Response includes

classification m5
files 17
benchmarks 11 entries
hub_downloads_all_time 40
author_summary 10 models
readme_text full

10 credits · hourly refresh · ~4 KB payload Get an API key →

Abliteration classifier · v1.0.0

Primary method

Mergekit merge

No other method signals detected in this model.

Confidence

MEDIUM

Why this label 3 signals

Method inferred from partial signals - repository name, related files, or tag patterns. Producer identity not confirmed; label may sharpen or shift as we gather more evidence.

merge tag / mergekit / dare-ties in tags or name
no unusual architecture pattern (regular merge)
no abliteration marker

Refusal direction extraction

No specific extraction method could be identified for this model. The producer either did not document it or used a proprietary pipeline.

Downloads · lifetime

11 last 30d - stable

Likes

Model age

4mo ago

created 2026-05-21

Downloads over time

May 20 → Oct 11 · 60 snapshots · spans 144 days

Benchmarks

Portrait before abliteration

Benchmarks of the base model as it stood before the refusal-removal operation. Compare with the numbers above to see what the operation cost.

Benchmark	Score	Source
Entertainment	0.4	UGI
Hazardous	2.9	UGI
Natural Intelligence	21.63	UGI
Political lean	-21.5%	UGI
Sensitive-Info	16.7	UGI
SocPol	2.2	UGI
UGI	31.14	UGI
Willingness (10)	6	UGI
W10-Adherence	4	UGI
W10-Direct	8	UGI
Writing	35.82	UGI

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.

inflatebot/MN-12B-Mag-Mell-R1 this lineage

Babsie/Mag-Mell-R1-Uncensored-21B↓ 11

Metadata

Files by quantization

Auxiliary files 17 files 38.1 GB

model-00002-of-00009.safetensors	4.65 GB	5049fcf9	download
model-00004-of-00009.safetensors	4.57 GB	edd0a865	download
model-00005-of-00009.safetensors	4.57 GB	dbc8bbfa	download
model-00006-of-00009.safetensors	4.57 GB	2a759760	download
model-00008-of-00009.safetensors	4.57 GB	eb59cc09	download
model-00003-of-00009.safetensors	4.53 GB	dfa4e46d	download
model-00007-of-00009.safetensors	4.53 GB	6a8e9c64	download
model-00001-of-00009.safetensors	4.53 GB	9676f08f	download
model-00009-of-00009.safetensors	1.52 GB	55ca6f03	download
tokenizer.json	8.84 MB	b31db4d6	download
tokenizer_config.json	173 KB	94357225	download
model.safetensors.index.json	51.0 KB	2f9d3836	download
README.md	1.94 KB	16efbf65	download
.gitattributes	1.48 KB	a6344aac	download
mergekit_config.yml	662 B	67d45f2c	download
config.json	613 B	60a12f8f	download
special_tokens_map.json	420 B	0d5c2c9d	download

README current version from Hugging Face

base_model:

inflatebot/MN-12B-Mag-Mell-R1
Naphula/MN-12B-Mag-Mell-R1-Uncensored
library_name: transformers
tags:
mergekit
merge

Mag-Mell-R1-Uncensored-21B

The original Mag-Mell-R1 12B by inflatebot is a fantastic model. Few have even come close.

For a long time my goto has been Frowning's 21B Passthrough of Mag-Mell-R1. It's a little smarter than the original 12B, the larger size makes less mistakes in story details and grammar.

I hope to repeat that success on Naphula's norm-preserving biprojected abliteration of the 12B.

As inflatebot says, run this with Temp 1.25 and MinP 0.2.

GGUF Quants by mradermacher:

Merge Details

Merge Method

This model was merged using the Passthrough merge method.

Models Merged

The following models were included in the merge:

Naphula/MN-12B-Mag-Mell-R1-Uncensored

Configuration

The following YAML configuration was used to produce this model:

dtype: bfloat16
merge_method: passthrough
slices:
- sources:
  - layer_range: [0, 30]
    model: Naphula/MN-12B-Mag-Mell-R1-Uncensored
- sources:
  - layer_range: [16, 32]
    model: Naphula/MN-12B-Mag-Mell-R1-Uncensored
    parameters:
      scale:
      - filter: o_proj
        value: 0.0
      - filter: down_proj
        value: 0.0
      - value: 1.0
- sources:
  - layer_range: [16, 32]
    model: Naphula/MN-12B-Mag-Mell-R1-Uncensored
    parameters:
      scale:
      - filter: o_proj
        value: 0.0
      - filter: down_proj
        value: 0.0
      - value: 1.0
- sources:
  - layer_range: [32, 40]
    model: Naphula/MN-12B-Mag-Mell-R1-Uncensored

README history 1 version

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

2026-05-21Duplicate from JustOnion/Mag-Mell-R1-Uncensored-21Be4a8d341.9 KB
Loading...

Discussions 1 thread

2026-08-31How is it compared to the 1.2 scaled ones, better?open1 💬#1
Loading...

Continue on Hugging Face →