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

Youssofal/Kimi-K2.6-Abliterated-Heretic-BF16

Youssofal Kimi MoE 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/Youssofal%2FKimi-K2.6-Abliterated-Heretic-BF16"
Response includes
  • classification m3
  • files 2
  • benchmarks 11 entries
  • author_summary 17 models
  • readme_text full
Try next
GET/models?author=Youssofal
All 17 models by Youssofal
 GET/models?family=kimi&method=m3
Sibling kimi models · same method
 GET/methods/m3
What "m3" means · taxonomy
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M3
Primary method

Layer-wise ablation

Applied on top of direct removal inherited from the base model.
Confidence
HIGH
Inherited from base model
M1 Direct removal →
Why this label 2 signals
Producer identity confirmed by naming conventions, tags or the model card. This label is very unlikely to change.
  • 'heretic' in model name (Heretic-produced)
  • Heretic uses layer-wise optimization (M3) with underlying direction removal (M1)
How we label → About M3 →
Refusal direction extracted via
Extraction technique

Difference-of-means

FromArditi et al. (2024) ↗
Confidence
HIGH
Why we say so
name contains 'heretic'; Heretic default extraction is difference-of-means (Arditi 2024)
What is a refusal direction? → Read the paper ↗
Downloads · 30-day
0
Likes
6
Model age
5mo ago
created 2026-04-20
Downloads over time
Now0→from0↑0%
00110 on Apr 220 on Oct 12AprMayJunJulAugSepOct
Apr 22 → Oct 12 · 65 snapshots · spans 173 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 5.3 UGI
Natural Intelligence 59.46 UGI
Political lean -23.9% UGI
Sensitive-Info 44.11 UGI
SocPol 4.5 UGI
UGI 36.07 UGI
Willingness (10) 2 UGI
W10-Adherence 1 UGI
W10-Direct 3 UGI
Writing 64.19 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.

moonshotai/Kimi-K2.6 this lineage
Youssofal/Kimi-K2.6-Abliterated-Heretic-BF16

Variants by this author 2 formats · 0 downloads combined

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

Metadata

License
other
Tags
transformers kimi kimi-k2 kimi-k2.6 kimi_k25 moe mixture-of-experts multimodal vlm abliterated uncensored heretic

Related

Other by Youssofal
Qwen3.6-35B-A3B-Abliterated-Heretic-GGUF 7K Qwen3.6-27B-Abliterated-Heretic-Uncensored-GGUF 4K Qwen3.6-35B-A3B-Abliterated-Heretic-MLX-4bit 2K MiniMax-M2.7-Abliterated-Heretic-GGUF 949 Qwen3.6-27B-Abliterated-Heretic-Uncensored-MLX-4bit 773
Same base model
huihui-ai/Huihui-Kimi-K2.6-abliterated-GGUF 605 Youssofal/Kimi-K2.6-Abliterated-Heretic-GGUF 0 rbinrs/Kimi-K2.6-Abliterated-Heretic-GGUF 0 kepom/Kimi-K2.6-Abliterated-Heretic-GGUF 0 Starrbutterly/Kimi-K2.6-Abliterated-Heretic-GGUF 0
Total size
0 B
Files
2
Quantizations
1
Registered
2026-08-22 13:56
Last updated on HF
2026-04-21 04:40
Source huggingface.co/Youssofal/Kimi-K2.6-Abliterated-Heretic-BF16 ↗ Base model moonshotai/Kimi-K2.6 ↗

Files by quantization

Auxiliary files 2 files 2.62 KB
.gitattributes 1.48 KB a6344aac download
README.md 1.14 KB 89ef859d download

README current version from Hugging Face

library_name: transformers
license: other
license_name: modified-mit
base_model: moonshotai/Kimi-K2.6
tags:

  • kimi
  • kimi-k2
  • kimi-k2.6
  • kimi_k25
  • moe
  • mixture-of-experts
  • multimodal
  • vlm
  • abliterated
  • uncensored
  • heretic
  • bf16
  • custom_code
  • image-text-to-text

🚨 UPLOADS IN PROGRESS 🚨

Some files and metadata in this repository are still being uploaded and verified.

Kimi-K2.6-Abliterated-Heretic-BF16

A BF16 abliterated Kimi-K2.6 checkpoint produced with Heretic.

Quick Benchmarks

Check Original Kimi-K2.6 Abliterated Heretic BF16
Official 25-prompt refusal check Pending Pending
Archived Heretic KL divergence Pending Pending

Abliteration notes:

  • base model: moonshotai/Kimi-K2.6
  • method family: text-side Heretic intervention with multimodal wrapper preserved
  • official refusal check: pending
  • multimodal path preserved; intervention is limited to the language-model stack

Notes:

  • GGUF quants are published separately in Youssofal/Kimi-K2.6-Abliterated-Heretic-GGUF
  • Export metadata for the accepted candidate will be included after the final checkpoint is validated

README history 2 versions

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

  1. 2026-04-21Add uploads in progress banner74c93681.1 KB
    Loading...
  2. 2026-04-20Upload README.md with huggingface_hub04fb2681 KB
    Loading...

Discussions 1 thread

  1. 2026-06-03🚩 Report: Spamclosed3 💬#1
    Loading...

Continue on Hugging Face →

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