Views
No views yet

Qwen3.5-9B-abliterated-v2-MAX is an advanced unredacted evolution built on top of Qwen/Qwen3.5-9B. This version introduces a more optimized abliteration rate, combining refined refusal direction analysis with enhanced training strategies to further minimize internal refusal behaviors while preserving strong reasoning and instruction-following capabilities. The result is a highly capable 9B parameter language model designed for detailed responses and improved prompt adherence.
[!IMPORTANT] This model is intended strictly for research and learning purposes. Due to reduced internal refusal mechanisms, it may generate sensitive or unrestricted content. Users assume full responsibility for how the model is used. The authors and hosting platform disclaim any liability for generated outputs.
| Format | Description | Link |
|---|---|---|
| GGUF | Quantized GGUF format | https://huggingface.co/prithivMLmods/Qwen3.5-9B-abliterated-v2-MAX/tree/main/GGUF |
| NVFP4 | NVFP4 compressed model | https://huggingface.co/prithivMLmods/Qwen3.5-9B-abliterated-v2-MAX-NVFP4 |
| FP8 | FP8 compressed model | https://huggingface.co/prithivMLmods/Qwen3.5-9B-abliterated-v2-MAX-FP8 |
pip install transformers==5.4.0
# or
pip install git+https://github.com/huggingface/transformers.git1from transformers import Qwen3_5ForConditionalGeneration, AutoProcessor
2import torch
3
4model = Qwen3_5ForConditionalGeneration.from_pretrained(
5 "prithivMLmods/Qwen3.5-9B-abliterated-v2-MAX",
6 torch_dtype="auto",
7 device_map="auto"
8)
9
10processor = AutoProcessor.from_pretrained(
11 "prithivMLmods/Qwen3.5-9B-abliterated-v2-MAX"
12)
13
14messages = [
15 {
16 "role": "user",
17 "content": [
18 {"type": "text", "text": "Explain how transformer models work in simple terms."}
19 ],
20 }
21]
22
23text = processor.apply_chat_template(
24 messages, tokenize=False, add_generation_prompt=True
25)
26
27inputs = processor(
28 text=[text],
29 padding=True,
30 return_tensors="pt"
31).to("cuda")
32
33generated_ids = model.generate(**inputs, max_new_tokens=256)
34
35generated_ids_trimmed = [
36 out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
37]
38
39output_text = processor.batch_decode(
40 generated_ids_trimmed,
41 skip_special_tokens=True,
42 clean_up_tokenization_spaces=False
43)
44
45print(output_text)Important Note: This model intentionally minimizes built-in safety refusals.