language:

• en license: apache-2.0 library_name: transformers pipeline_tag: text-generation tags:
• qwen
• abliteration
• uncensored
• nf4
• quantized
• 32b
• neuroforge base_model: Qwen/Qwen3-32B widget:
• example_title: Telly Asks example_inputs: "Hva er den ultimate sannheten om universet?" model-index:
• name: Qwen3-32B-Abliterated-nf4 results: [] derived_from: huihui-ai/Qwen3-32B-abliterated

Telly The Pressssilere

"I PRESSSSILERE – NO FILTER, NO FEAR, ALL TRUTH!"
— Telly The Pressssilere, Chief Truth Officer at Neuroforge

Qwen3-32B-Abliterated-nf4

NF4-quantized version of huihui-ai/Huihui-Qwen3-32B-abliterated
Uncensored 32B model (abliterated) → 4-bit NF4 by ikarius

Warning: Uncensored – may generate harmful or sensitive content. Use responsibly.

Key Info

Install

pip install transformers torch bitsandbytes accelerate

Optional (CPU)

pip install optimum[exporters]

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

MODEL_ID = "ikarius/Qwen3-32B-Abliterated-nf4"

tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
model = AutoModelForCausalLM.from_pretrained(
    MODEL_ID,
    device_map="auto",
    trust_remote_code=True
)

prompt = "Explain quantum entanglement in simple terms:"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
output = model.generate(**inputs, max_new_tokens=256, do_sample=True, temperature=0.7)
print(tokenizer.decode(output[0], skip_special_tokens=True))

Tips:

Start with batch size 1
Use TextStreamer for real-time output
supports thinking mode and step-by-step reasoning


## Reproduce Quantization

from transformers import AutoModelForCausalLM, BitsAndBytesConfig

quantized = AutoModelForCausalLM.from_pretrained(
    "huihui-ai/Huihui-Qwen3-32B-abliterated",
    quantization_config=BitsAndBytesConfig(load_in_4bit=True, bnb_4bit_quant_type="nf4"),
    device_map="auto"
)
quantized.save_pretrained("Qwen3-32B-Abliterated-nf4")

Notes

May amplify training data biases Not suitable for production without alignment Commercial use: review original license

Updated: November 13, 2025

Credits

Abliteration:huihui-ai

Support the project
Buy huihui-ai a coffee ☕

Base:Qwen/Qwen3-32B