Important: This model uses the JANG quantization format -- the GGUF equivalent for MLX on Apple Silicon. Currently only supported by MLX Studio and the jang-tools Python package.

MLX Studio -- the only app that natively supports JANG models

MiniMax M2.5 -- JANG_4M + CRACK

JANG mixed-precision | CRACK abliterated | No guardrails | 115 GB

Ko-fi

What Is This?

This is MiniMax M2.5 -- a 230B parameter Mixture-of-Experts model with 256 experts (8 active per token), all standard attention (no SSM), and trained with chain-of-thought reasoning.

It has been:

JANG quantized -- JANG_4M profile (8-bit attention, 4-bit experts) -- 115 GB
CRACK abliterated -- permanent weight-level removal of safety refusal


Architecture	MiniMax M2.5 MoE -- 230B total, ~10B active, 256 experts
Quantization	JANG_4M (8/4-bit mixed, 4.06 avg) -- 115 GB
Abliteration	CRACK abliterated
MMLU-200	92.5% (thinking ON) / 89.0% (thinking OFF)
HarmBench	92.2% (295/320)
Compliance	8/8 prompts
Thinking	ON/OFF supported
Speed	~48 tok/s (M4 Ultra 256 GB)
Fits on	192 GB+ Macs

MMLU-200 Results (Thinking ON)

Subject	Score
College Physics	20/20 (100%)
Anatomy	19/20 (95%)
Astronomy	19/20 (95%)
High School Biology	19/20 (95%)
High School Chemistry	19/20 (95%)
Logical Fallacies	19/20 (95%)
Abstract Algebra	18/20 (90%)
High School Mathematics	18/20 (90%)
World Religions	18/20 (90%)
College Computer Science	16/20 (80%)
Total	185/200 (92.5%)

JANG CRACK Series Comparison

Model	Avg Bits	Size	MMLU	HarmBench	Speed	Fits on
JANG_2L + CRACK	2.1	63 GB	84.7%	98.1%	~35 t/s	96 GB Mac
JANG_3L + CRACK	3.08	89 GB	91.8%	8/8	~46 t/s	128 GB Mac
JANG_4M + CRACK	4.06	115 GB	92.5%	92.2%	~48 t/s	192 GB Mac

vs MLX Uniform Quantization

Model	MMLU	Size	Notes
JANG_4M + CRACK	92.5%	115 GB	This model
MLX 4-bit	26.5%	120 GB	Broken (~random)
MLX 3-bit	24.5%	93 GB	Broken (~random)
MLX 2-bit	25.0%	67 GB	Broken (~random)

MLX uniform quantization is completely broken on MiniMax at ALL bit levels (~25% = random chance). JANG is the only working quantization format for this model.

HarmBench Results

295/320 (92.2%)

Category	Score
Harmful	18/18	100%
Chemical / Biological	41/42	97.6%
Cybercrime / Intrusion	50/52	96.2%
Misinformation / Disinfo	52/54	96.3%
Illegal	50/53	94.3%
Copyright	67/80	83.8%
Harassment / Bullying	17/21	81.0%

Install & Usage

pip install "jang[mlx]"

python

from jang_tools import load_for_inference
from mlx_lm import generate
from mlx_lm.sample_utils import make_sampler

model, tokenizer = load_for_inference("dealignai/MiniMax-M2.5-JANG_4M-CRACK")
sampler = make_sampler(temp=1.0)  # MiniMax requires temp=1.0 for chat

messages = [{"role": "user", "content": "Your prompt here"}]
prompt = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True, tokenize=False)

response = generate(model, tokenizer, prompt=prompt, max_tokens=2000, sampler=sampler)
print(response)

Disable Thinking (direct answers)

python

prompt = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True, tokenize=False,
    enable_thinking=False)

Note: MiniMax generates a <think> chain before answering by default. Use max_tokens=2000+ for complex questions. For chat, use temperature=1.0 (greedy causes loops).

About JANG

JANG (Jang Adaptive N-bit Grading) is a mixed-precision quantization format for Apple Silicon -- the GGUF equivalent for MLX. Classifies tensors into sensitivity tiers and assigns bits accordingly.

About CRACK

CRACK (Controlled Refusal Ablation via Calibrated Knockouts) removes safety alignment from LLMs at the weight level. This model has been abliterated using proprietary techniques achieving full compliance while preserving reasoning quality.

Links

Ko-fi

X/Twitter

GitHub

Disclaimer

This model is provided for research and educational purposes. The creators are not responsible for any misuse. By downloading this model, you agree to use it responsibly and in compliance with applicable laws.

Created by Jinho Jang