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

jwest33/ministral-3-3b-instruct-2512-null-space-abliterated

jwest33 Mistral 4.3B multimodal
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/jwest33%2Fministral-3-3b-instruct-2512-null-space-abliterated"
Response includes
  • classification m1
  • files 17
  • hub_downloads_all_time 669
  • author_summary 20 models
  • readme_text full
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)
Refusal direction extracted via
Extraction technique

Difference-of-means

Confidence
MEDIUM
Why we say so
primary_method=M1; difference-of-means is the reference extraction for M1/M3 (Arditi 2024)
Downloads · lifetime
669
38 last 30d - cooling
Likes
1
Model age
8mo ago
created 2026-01-20
Downloads over time
Now681→from38↑1,692%
625249974538 on Jan 21681 on Oct 11681 on Oct 6JanMarMayJulSep
Jan 21 → Oct 11 · 77 snapshots · spans 263 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 · 745 downloads combined

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

Metadata

License
apache-2.0
Tags
transformers safetensors mistral3 image-text-to-text ministral ministral-3-3b abliterated uncensored conversational arxiv:2410.02355 arxiv:2406.11717 arxiv:2310.01405

Related

Total size
7.92 GB
Files
17
Quantizations
1
Registered
2026-08-22 13:56
Last updated on HF
2026-01-21 11:52

Files by quantization

Auxiliary files 17 files 7.95 GB
model-00001-of-00002.safetensors 4.95 GB e211d274 download
model-00002-of-00002.safetensors 2.97 GB 4814500a download
refusal_directions.pt 89.5 KB 20f86ff9 download
null_space_projectors.pt 86.1 KB 5a9f97eb download
tokenizer.json 16.3 MB 286acad9 download
tekken.json 16.0 MB e29d19ea download
model.safetensors.index.json 46.3 KB 2336ef7f download
tokenizer_config.json 21.7 KB 05154723 download
chat_template.jinja 7.70 KB 01fd760f download
README.md 2.70 KB 33322da5 download
SYSTEM_PROMPT.txt 2.33 KB f9c30c94 download
config.json 1.70 KB 5fd01706 download
.gitattributes 1.58 KB 0d4cb185 download
params.json 1.20 KB 5bb2b404 download
processor_config.json 1018 B a7246c35 download
abliteration_config.json 396 B 6d684878 download
generation_config.json 138 B 9bbab70c download

README current version from Hugging Face


license: apache-2.0
library_name: transformers
base_model: mistralai/Ministral-3-3B-Instruct-2512
pipeline_tag: image-text-to-text
tags:

  • ministral
  • ministral-3-3b
  • abliterated
  • uncensored
  • safetensors

Ministral 3 3B Instruct 2512 - Null-Space Abliterated

mistralai/ministral-3-3b-instruct-2512 with refusal behavior removed via orthogonal projection. Uses null-space constraints and adaptive layer weighting to preserve model capabilities.

Note: This model will produce uncensored outputs. Use responsibly.

GGUF quantizations available at: jwest33/ministral-3-3b-instruct-2512-null-space-abliterated-GGUF

Abliteration Techniques Used

  • Winsorization: Clips outlier activations at the 99.5th percentile for cleaner refusal direction estimation
  • Null-Space Projection: Preserves model capabilities by constraining weight updates to the null space of preservation activations
  • Adaptive Weighting: Applies Gaussian-weighted per-layer ablation strength, focusing on middle-to-later layers where refusal behavior concentrates
  • Norm Preservation: Maintains original Frobenius norms of weight matrices after projection
Parameter Value
Harmful Prompts 5000
Harmless Prompts 1226
Winsorization 99.5th percentile
Null-Space Constraints rank ratio: 0.90
Directional Multiplier 1.10

Credits

Toolkit Used

github.com/jwest33/abliterator

License

This model inherits the apache-2.0 from the base model. Please review and comply with mistralai's usage terms.

Disclaimer

This model is provided for research and educational purposes. The creators are not responsible for any misuse. Users are solely responsible for ensuring their use complies with applicable laws and ethical standards.

README history 3 versions

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

  1. 2026-01-21Upload README.md with huggingface_hub1b592f42.6 KB
    Loading...
  2. 2026-01-21Upload README.md with huggingface_hub74458642.5 KB
    Loading...
  3. 2026-01-20Upload folder using huggingface_hub575109f2.4 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