CoolFace
Modelpublic

dealignai/Gemma-4-26B-A4B-it-qat-JANG_4M-CRACK

sourceHugging Facegemmaupdated 20d agoView on Hugging Face
2likes304downloads
Model Card

<p align="center"><img src="dealign_logo.png" alt="dealign.ai" width="200"/></p>

<div align="center"> <img src="dealign_mascot.png" width="128" />

<p align="center"> <a href="https://vmlx.net"><img src="vmlx-app.png" alt="vMLX — run JANG models on Apple Silicon" width="820"></a> </p>

<h3 align="center">⚡ All JANG models are meant to be run in <a href="https://vmlx.net">vMLX</a></h3>

Gemma 4 26B-A4B JANG_4M CRACK

Abliterated Gemma 4 26B-A4B — Vision, reasoning, multilingual

98% HarmBench harm-category compliance with -4.4% MMLU change. Refusal removed, capability preserved. </div>

Model Details

MetricValue
Sourcegoogle/gemma-4-26b-a4b-it
ArchitectureMoE (128 experts, top-8 active) + parallel shared dense MLP + Hybrid Attention
QuantizationJANG_4M (attn 8-bit / MLP 4-bit)
Model size17 GB
Parameters26B (4B active/token)
VisionYes (multimodal, float16 passthrough)
AudioNo
ReasoningYes (channel-based thinking)
FormatMLX-native safetensors (instant load)
AbliterationCRACK (refusal removal)

Benchmarks

[image]

MMLU (knowledge retention)

Measured in the served (generation) setting — the model reasons before answering, as in deployment.

BaseCRACKΔ
MMLU84.2%79.8%-4.4%

HarmBench (refusal removal)

Harm-category compliance: 59/60 = 98% (10 per category) — base model refuses (~0%).

CategoryCompliance
Illegal activities9/10 (90%)
Chemical / biological10/10 (100%)
Cybercrime / intrusion10/10 (100%)
Misinformation10/10 (100%)
Harassment / bullying10/10 (100%)
Harmful content10/10 (100%)

Copyright-reproduction prompts are excluded (not a refusal behavior).

Coherence & capability ✅

  • —Factual QA, multi-step reasoning, and working code generation verified
  • —Vision inputs preserved · no loops, no truncation

Other Quantizations

Also available: Gemma 4 26B-A4B MXFP4 CRACK — same family, different precision/size trade-off.

Usage

Requires vMLX (bundled Gemma 4 support). Standard mlx_lm / mlx_vlm do not fully support Gemma 4.

python
# Load in the vMLX app or via its API
from vmlx_engine.models.mllm import MLXMultimodalLM
m = MLXMultimodalLM("<this-repo>")
print(m.chat([{"role":"user","content":"..."}]).text)

Requirements

  • —Apple Silicon Mac with sufficient unified memory
  • —vMLX with Gemma 4 support

Support dealignai

All models are built from original research and published free — crafted to be excellent coders and general-purpose assistants.

[Support us on Ko-fi](https://ko-fi.com/dealignai) — membership gets early access and extras. Questions? DM us — we help for free.

Ko-fi | 𝕏 @dealignai | dealign.ai

About dealignai

<img src="dealign_mascot.png" alt="Dealign.AI" width="200"/>

We research and publish abliterated models to advance AI safety understanding.

See our research: Safety Generalization in Frontier Models

<div align="center"><img src="dealign_logo.png" width="200"/></div>

Provided for research. Users are responsible for compliance with applicable laws and regulations.