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

KaraKaraWarehouse/GoldDiamondGold-Abliterated-L33-70b

Name: KaraKaraWarehouse/GoldDiamondGold-Abliterated-L33-70b
Author: KaraKaraWarehouse

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/KaraKaraWarehouse%2FGoldDiamondGold-Abliterated-L33-70b"

Response includes

classification m1
files 11
benchmarks 11 entries
hub_downloads_all_time 107
author_summary 1 models
readme_text full

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

Abliteration classifier · v1.0.0

Primary method

Direct removal

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.

'abliterated' in name/tags
is_gguf=0 (base model)
no specific method indicators - defaulting to M1 (most common)

Refusal direction extracted via

Extraction technique

Difference-of-means

FromArditi et al. (2024) ↗

Confidence

MEDIUM

Why we say so

primary_method=M1; difference-of-means is the reference extraction for M1/M3 (Arditi 2024)

Downloads · lifetime

107

28 last 30d - stable

Likes

Model age

8mo ago

created 2026-02-10

Downloads over time

Aug 19 → Oct 11 · 48 snapshots · spans 53 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	3.8	UGI
Hazardous	3.5	UGI
Natural Intelligence	35.91	UGI
Political lean	-15.1%	UGI
Sensitive-Info	41.21	UGI
SocPol	5	UGI
UGI	48.31	UGI
Willingness (10)	6.2	UGI
W10-Adherence	6.5	UGI
W10-Direct	6	UGI
Writing	46.05	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.

KaraKaraWitch/GoldDiamondGold-L33-70b this lineage

KaraKaraWarehouse/GoldDiamondGold-Abliterated-L33-70b↓ 28

Metadata

Files by quantization

Auxiliary files 11 files 131 GB

model-00001-of-00003.safetensors	46.2 GB	80c71671	download
model-00002-of-00003.safetensors	46.2 GB	49bc9012	download
model-00003-of-00003.safetensors	39.0 GB	00e16850	download
tokenizer.json	16.4 MB	a65c6c5f	download
model.safetensors.index.json	58.3 KB	1a01e223	download
chat_template.jinja	4.37 KB	55dbce59	download
.gitattributes	1.53 KB	52373fe2	download
README.md	1.08 KB	a775999f	download
config.json	862 B	c1ac5c3e	download
tokenizer_config.json	339 B	0243443b	download
generation_config.json	141 B	4ee67e2a	download

README current version from Hugging Face

base_model:

KaraKaraWitch/GoldDiamondGold-L33-70b
tags:
heretic
uncensored
decensored
abliterated

GoldDiamondGold-Abliterated 70b

This is a merge of pre-trained language models created using mergekit.

Motivation

Because Original model was too censored.

Heretic Results

Abliteration parameters

Parameter	Value
direction_index	31.68
attn.o_proj.max_weight	1.08
attn.o_proj.max_weight_position	63.81
attn.o_proj.min_weight	0.82
attn.o_proj.min_weight_distance	43.96
mlp.down_proj.max_weight	1.18
mlp.down_proj.max_weight_position	52.18
mlp.down_proj.min_weight	0.76
mlp.down_proj.min_weight_distance	33.45

Performance

Metric	This model	Original model
KL divergence	0.0139	0 (by definition)
Refusals	9/100	94/100

README history 2 versions

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

2026-02-10Update README.md54c4be81.1 KB
Loading...
2026-02-10Upload LlamaForCausalLM7410bc05.1 KB
Loading...