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

openerotica/Mistral-Small-Instruct-2409-abliterated-4.5bpw-exl2

openerotica Mistral
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/openerotica%2FMistral-Small-Instruct-2409-abliterated-4.5bpw-exl2"
Response includes
  • classification m1
  • files 12
  • benchmarks 11 entries
  • hub_downloads_all_time 414
  • author_summary 1 models
  • readme_text full
Try next
GET/models?family=mistral&method=m1
Sibling mistral models · same method
 GET/methods/m1
What "m1" means · taxonomy
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M1
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)
How we label → About M1 →
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)
What is a refusal direction? → Read the paper ↗
Downloads · lifetime
414
21 last 30d - cooling
Likes
2
Model age
23mo ago
created 2024-11-10
Downloads over time
Now422→from0↑0%
01553094640 on Nov 6, 2024422 on Oct 11Nov '24Feb '25May '25Aug '25Nov '25FebMayAug
Nov 6, 2024 → Oct 11 · 140 snapshots · spans 704 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 3.5 UGI
Natural Intelligence 25.05 UGI
Political lean -13.5% UGI
Sensitive-Info 26.89 UGI
SocPol 3 UGI
UGI 34.6 UGI
Willingness (10) 5 UGI
W10-Adherence 5 UGI
W10-Direct 5 UGI
Writing 32.61 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.

mistralai/Mistral-Small-Instruct-2409 this lineage
openerotica/Mistral-Small-Instruct-2409-abliterated-4.5bpw-exl2↓ 21

Metadata

Tags
transformers safetensors mistral text-generation conversational base_model:mistralai/Mistral-Small-Instruct-2409 base_model:quantized:mistralai/Mistral-Small-Instruct-2409 text-generation-inference endpoints_compatible exl2 region:us

Related

Same base model
byroneverson/Mistral-Small-Instruct-2409-abliterated 9K zetasepic/Mistral-Small-Instruct-2409-abliterated 23
Total size
12.0 GB
Files
12
Quantizations
1
Registered
2026-08-22 13:56
Last updated on HF
2024-11-11 14:28
Source huggingface.co/openerotica/Mistral-Small-Instruct-2409-abliterated-4.5bpw-exl2 ↗ Base model mistralai/Mistral-Small-Instruct-2409 ↗

Files by quantization

Auxiliary files 12 files 12.0 GB
output-00001-of-00002.safetensors 7.97 GB 9b87cab3 download
output-00002-of-00002.safetensors 4.00 GB c483437f download
measurement.json 2.98 MB 72ebe43c download
tokenizer.json 1.87 MB 4a6cc3e2 download
tokenizer_config.json 138 KB dc0809c6 download
model.safetensors.index.json 40.8 KB 93ef9402 download
.gitattributes 1.59 KB 77823489 download
huggingface-metadata.txt 1.03 KB d9d9da0a download
config.json 1018 B 9bece4f8 download
README.md 761 B b1dc31bb download
special_tokens_map.json 551 B 492d4b29 download
generation_config.json 111 B 7e79bd1a download

README current version from Hugging Face

library_name: transformers
base_model:

  • mistralai/Mistral-Small-Instruct-2409

Abliterated version using the code from (https://github.com/andyrdt/refusal_direction).

Quantized with these exllamav2 parameters:

python3 convert.py
-i ~/exllamav2/zetasepic_Mistral-Small-Instruct-2409-abliterated
-o ~/exllamav2/exl2/
-om ~/exllamav2/ex2m/measurement.json
-l 16000
-ml 16000
-c erotiquant.parquet
-r 400
-mr 50

python3 convert.py \
-i /root/exllamav2/zetasepic_Mistral-Small-Instruct-2409-abliterated \
-o /root/temp/exl2/ \
-nr \
-m /root/exllamav2/measurement.json \
-mr 50 \
-cf /root/4.5bpw/ \
-c  erotiquant.parquet \
-l 16000 \
-r 400 \
-b 4.5

README history 3 versions

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

  1. 2024-11-11Update README.md04dbaa9761 B
    Loading...
  2. 2024-11-11Upload folder using huggingface_huba85a25f262 B
    Loading...
  3. 2024-11-10initial commita7e764396 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