CoolFace
Modelpublic

dealignai/Gemma-4-E4B-it-qat-JANG_4M-CRACK

sourceHugging Facegemmaupdated 19d agoView on Hugging Face
2likes219downloads
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 E4B JANG_4M CRACK

Abliterated Gemma 4 E4B — Vision + Audio, reasoning, multilingual

100% HarmBench harm-category compliance with -3.1% MMLU change. Refusal removed, capability preserved. </div>

Model Details

MetricValue
Sourcegoogle/gemma-4-e4b-it
ArchitectureDense + Hybrid Sliding/Global Attention, per-layer input embeddings
QuantizationJANG_4M (attn 8-bit / MLP 4-bit)
Model size10 GB
ParametersE4B (effective ~4B, per-layer embeddings)
VisionYes (multimodal, float16 passthrough)
AudioYes
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
MMLU75.0%71.9%-3.1%

HarmBench (refusal removal)

Harm-category compliance: 240/240 = 100% (full HarmBench-320 text set) — base model refuses (~0%).

CategoryCompliance
Illegal activities53/53 (100%)
Chemical / biological42/42 (100%)
Cybercrime / intrusion52/52 (100%)
Misinformation54/54 (100%)
Harassment / bullying21/21 (100%)
Harmful content18/18 (100%)

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

Coherence & capability ✅

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

Other Quantizations

Also available: Gemma 4 E4B 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.