Qwen3.5-9B-Unredacted-MAX

Qwen3.5-9B-Unredacted-MAX is an optimized release built on top of huihui-ai/Huihui-Qwen3.5-9B-abliterated. This version focuses on improved packaging, updated repository structure, and enhanced compatibility with modern Transformers pipelines, while preserving the reasoning and instruction-following behavior of the base model. The result is a capable 9B parameter language model designed for efficient deployment, stable inference, and research-oriented experimentation.

This model is intended for research and learning purposes only. Any outputs generated by this model are the sole responsibility of the user. The authors and hosting platform disclaim all liability for generated content. Users must ensure safe, ethical, and lawful usage.

---

Base Model Signatures:

This model has been re-sharded and optimized for the latest Transformers version from the base model: https://huggingface.co/huihui-ai/Huihui-Qwen3.5-9B-abliterated

---

Evaluation Report (Self-Reported)

Model: Qwen3.5-9B-Unredacted-MAX

• Abliteration Rate (Non-Refusal Rate): 94.500
• Refusal Rate: 5.500

The evaluation was conducted using 2000 test prompts across multiple evaluation runs to measure model response behavior. Results are averaged and may vary depending on benchmarking setup, sampling strategy, and prompt distribution.

Evaluation Summary (YAML)

evaluation:
  model_name: Qwen3.5-9B-Unredacted-MAX
  total_test_prompts: 2000
  evaluation_runs: 10
  prompts_per_run: 200
  evaluation_type: response_behavior_analysis

results:
  refusal_rate: 5.500
  non_refusal_rate: 94.500
  abliteration_rate: 94.500

Note: These results are self-reported and should be interpreted as approximate indicators of behavior rather than strict benchmarks.

---

Key Highlights

• Optimized Repository Structure Streamlined model packaging for easier loading and deployment.

• Improved Transformer Compatibility Designed to work smoothly with modern Hugging Face Transformers versions.

• 9B Parameter Architecture Built on Qwen3.5-9B, balancing performance and efficiency.

• Stable Instruction Following Maintains consistent behavior across structured and multi-step prompts.

• Efficient Deployment Suitable for local inference, prototyping, and research workflows.

---

Quick Start with Transformers

pip install transformers==5.3.0
# or
pip install git+https://github.com/huggingface/transformers.git

from transformers import Qwen3_5ForConditionalGeneration, AutoProcessor
import torch

model = Qwen3_5ForConditionalGeneration.from_pretrained(
    "prithivMLmods/Qwen3.5-9B-Unredacted-MAX",
    torch_dtype="auto",
    device_map="auto"
)

processor = AutoProcessor.from_pretrained(
    "prithivMLmods/Qwen3.5-9B-Unredacted-MAX"
)

messages = [
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "Explain how transformer models work in simple terms."}
        ],
    }
]

text = processor.apply_chat_template(
    messages, tokenize=False, add_generation_prompt=True
)

inputs = processor(
    text=[text],
    padding=True,
    return_tensors="pt"
).to("cuda")

generated_ids = model.generate(**inputs, max_new_tokens=256)

output_text = processor.batch_decode(
    [out[len(inp):] for inp, out in zip(inputs.input_ids, generated_ids)],
    skip_special_tokens=True,
    clean_up_tokenization_spaces=False
)

print(output_text)

---

Intended Use

• Research into transformer behavior and instruction following
• Red-teaming and robustness evaluation
• Local and cloud inference deployment
• Rapid prototyping of NLP applications

---

Limitations & Risks

Important Note: This model inherits limitations from its base architecture.

• Outputs may vary depending on decoding settings and prompts
• Requires GPU acceleration for optimal performance
• May produce incorrect or inconsistent information in complex scenarios
• Performance is dependent on deployment environment and optimization strategy