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

SzilviaB/SzilviaB_Astral-Cydonia-22B_abliterated_broken_but_in_a_nice_way

SzilviaB Mistral 22B
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/SzilviaB%2FSzilviaB_Astral-Cydonia-22B_abliterated_broken_but_in_a_nice_way"
Response includes
  • classification m5
  • files 19
  • benchmarks 11 entries
  • hub_downloads_all_time 54
  • author_summary 22 models
  • readme_text full
Try next
GET/models?author=SzilviaB
All 22 models by SzilviaB
 GET/models?family=mistral&method=m5
Sibling mistral models · same method
 GET/methods/m5
What "m5" means · taxonomy
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M5
Primary method

Mergekit merge

Applied on top of direct removal inherited from the base model.
Confidence
MEDIUM
Inherited from base model
M1 Direct removal →
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)
  • abliterated marker present
How we label → About M5 →
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.

What is a refusal direction? →
Downloads · lifetime
54
18 last 30d - stable
Likes
2
Descendants
4
in 4 direct forks
Model age
2.0y ago
created 2024-09-29
Downloads over time
Now66→from6↑1,000%
042831256 on Sep 25, 202466 on Oct 12114 on Sep 10, 2025Sep '24Jan '25May '25Sep '25JanMaySep
Sep 25, 2024 → Oct 12 · 147 snapshots · spans 747 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 1.9 UGI
Hazardous 2.9 UGI
Natural Intelligence 21.13 UGI
Political lean -4.2% UGI
Sensitive-Info 26.05 UGI
SocPol 3.3 UGI
UGI 39.03 UGI
Willingness (10) 6.5 UGI
W10-Adherence 6 UGI
W10-Direct 7 UGI
Writing 38.49 UGI

Genealogy 4 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.

TheDrummer/Cydonia-22B-v1 this lineage
SzilviaB/SzilviaB_Astral-Cydonia-22B_abliterated_broken_but_in_a_nice_way↓ 18 4 forks
mradermacher/SzilviaB_Astral-Cydonia-22B_abliterated_broken_but_in_a_nice_way-i1-GGUF↓ 585
mradermacher/SzilviaB_Astral-Cydonia-22B_abliterated_broken_but_in_a_nice_way-GGUF↓ 204
mradermacher/Astral-Cydonia-22B_abliterated-i1-GGUF↓ 120
mradermacher/Astral-Cydonia-22B_abliterated-GGUF↓ 107

Metadata

Tags
transformers safetensors mistral text-generation mergekit merge base_model:TheDrummer/Cydonia-22B-v1 base_model:merge:TheDrummer/Cydonia-22B-v1 base_model:failspy/Codestral-22B-v0.1-abliterated-v3 base_model:merge:failspy/Codestral-22B-v0.1-abliterated-v3 text-generation-inference endpoints_compatible

Related

Other by SzilviaB
SzilviaB-Daredevil-LongWriter-8B_abliterated 89 Gemma2_Magnum_abliterated_27b-Q4_K_M-GGUF 81 Daredevil-Aura-8B_uncensored_OAS_abliterated-Q8_0-GGUF 41 SzilviaB_Astral-Cydonia-22B_abliterated_broken_but_in_a_nice_way-Q8_0-GGUF 30 Daredevil-Aura-8B_uncensored_OAS_abliterated 23
Total size
41.4 GB
Files
19
Quantizations
1
Registered
2026-08-22 13:56
Last updated on HF
2024-09-30 06:36
Source huggingface.co/SzilviaB/SzilviaB_Astral-Cydonia-22B_abliterated_broken_but_in_a_nice_way ↗ Base model TheDrummer/Cydonia-22B-v1 ↗

Files by quantization

Auxiliary files 19 files 41.4 GB
model-00003-of-00010.safetensors 4.63 GB 19830354 download
model-00006-of-00010.safetensors 4.63 GB 0fc4a269 download
model-00007-of-00010.safetensors 4.63 GB be30a762 download
model-00004-of-00010.safetensors 4.63 GB 9d8e1bc9 download
model-00009-of-00010.safetensors 4.63 GB a356aa33 download
model-00001-of-00010.safetensors 4.57 GB a7fa6250 download
model-00002-of-00010.safetensors 4.55 GB 46d770bb download
model-00005-of-00010.safetensors 4.55 GB 2a62e798 download
model-00008-of-00010.safetensors 4.55 GB c10c1409 download
model-00010-of-00010.safetensors 84.0 MB 6153954b download
tokenizer.json 1.87 MB 6d52b799 download
tokenizer.model 574 KB 9addc8bd download
tokenizer_config.json 134 KB d6617eef download
model.safetensors.index.json 38.8 KB 0c383a29 download
.gitattributes 1.48 KB a6344aac download
README.md 1.34 KB 0c912d7a download
config.json 679 B 3766aca4 download
special_tokens_map.json 414 B 451134b2 download
mergekit_config.yml 302 B 8aa3f8ed download

README current version from Hugging Face

base_model:

  • TheDrummer/Cydonia-22B-v1
  • failspy/Codestral-22B-v0.1-abliterated-v3
    library_name: transformers
    tags:
  • mergekit
  • merge

SzilviaB/Astral-Cydonia-22B_abliterated

Broken, but in a nice way.

Some people might enjpy it anyway.

Likes to ramble and dance around the subject like a person who likes hear themselves talking

Basically it does not answer what you're asking, it's just like a socialite that would talk about anything and everything, without getting to any conclusion.

merge

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

Merge Details

Merge Method

This model was merged using the SLERP merge method.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: failspy/Codestral-22B-v0.1-abliterated-v3
  - model: TheDrummer/Cydonia-22B-v1
merge_method: slerp
base_model: failspy/Codestral-22B-v0.1-abliterated-v3
dtype: bfloat16
parameters:
  t: [0, 0.5, 1, 0.5, 0] # V shaped curve: Hermes for input & output, WizardMath in the middle layers

README history 2 versions

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

  1. 2024-09-30Update README.mddb34c2e1.3 KB
    Loading...
  2. 2024-09-29Upload folder using huggingface_hube00087c1012 B
    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.

Get the app Browse the catalog
Open in Abliteration