This model is a int4 weight-only quantization with group_size 128 and symmetric quantization of DavidAU/Qwen3.5-21B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking generated by AutoRound. Please follow the license of the original model.

Quantization Details

Attribute	Value
Base Model	DavidAU/Qwen3.5-21B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking
Quantization Tool	AutoRound
Quantization Scheme	W4A16
Quantized Size	14806 MB

Evaluation Results

Task	Accuracy
hellaswag	0.5483
mmlu	0.6436
mmlu_abstract_algebra	0.4500
mmlu_anatomy	0.6889
mmlu_astronomy	0.8092
mmlu_business_ethics	0.7300
mmlu_clinical_knowledge	0.7283
mmlu_college_biology	0.7917
mmlu_college_chemistry	0.5700
mmlu_college_computer_science	0.5300
mmlu_college_mathematics	0.4000
mmlu_college_medicine	0.6301
mmlu_college_physics	0.5392
mmlu_computer_security	0.6900
mmlu_conceptual_physics	0.7617
mmlu_econometrics	0.4386
mmlu_electrical_engineering	0.7241
mmlu_elementary_mathematics	0.6587
mmlu_formal_logic	0.5714
mmlu_global_facts	0.4700
mmlu_high_school_biology	0.8097
mmlu_high_school_chemistry	0.6700
mmlu_high_school_computer_science	0.7200
mmlu_high_school_european_history	0.6909
mmlu_high_school_geography	0.7475
mmlu_high_school_government_and_politics	0.7772
mmlu_high_school_macroeconomics	0.7410
mmlu_high_school_mathematics	0.3889
mmlu_high_school_microeconomics	0.7605
mmlu_high_school_physics	0.6358
mmlu_high_school_psychology	0.8422
mmlu_high_school_statistics	0.6343
mmlu_high_school_us_history	0.6912
mmlu_high_school_world_history	0.7679
mmlu_human_aging	0.6592
mmlu_human_sexuality	0.7634
mmlu_humanities	0.5579
mmlu_international_law	0.7438
mmlu_jurisprudence	0.7407
mmlu_logical_fallacies	0.6871
mmlu_machine_learning	0.4643
mmlu_management	0.7961
mmlu_marketing	0.8291
mmlu_medical_genetics	0.6500
mmlu_miscellaneous	0.7854
mmlu_moral_disputes	0.6445
mmlu_moral_scenarios	0.3408
mmlu_nutrition	0.7222
mmlu_other	0.6971
mmlu_philosophy	0.7363
mmlu_prehistory	0.6728
mmlu_professional_accounting	0.4433
mmlu_professional_law	0.4668
mmlu_professional_medicine	0.7941
mmlu_professional_psychology	0.6095
mmlu_public_relations	0.5909
mmlu_security_studies	0.6816
mmlu_social_sciences	0.7199
mmlu_sociology	0.7662
mmlu_stem	0.6441
mmlu_us_foreign_policy	0.7900
mmlu_virology	0.4759
mmlu_world_religions	0.8363
piqa	0.7639

How to Use

HF Usage

Step 1: Install AutoRound

pip install auto-round

Step 2: Load and run the quantized model

python

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "Qwen3.5-21B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking-AutoRound-W4A16-RTN"

# load the tokenizer and the model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype="auto", device_map="auto")

# prepare the model input
prompt = "Write a quick sort algorithm."
messages = [{"role": "user", "content": prompt}]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

# conduct text completion
generated_ids = model.generate(**model_inputs, max_new_tokens=512)
output_ids = generated_ids[0][len(model_inputs.input_ids[0]) :].tolist()

content = tokenizer.decode(output_ids, skip_special_tokens=True)
print("content:", content)

VLLM Usage

bash

vllm serve Qwen3.5-21B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking-AutoRound-W4A16-RTN \
    --trust-remote-code \
    --dtype bfloat16 \
    --tensor_parallel_size 1

If you encounter any issues, feel free to open an issue on the AutoRound GitHub repo or provide feedback on the Low-Bit Open LLM Leaderboard.

Ethical Considerations and Limitations

The model can produce factually incorrect output, and should not be relied on to produce factually accurate information. Because of the limitations of the pretrained model and the finetuning datasets, it is possible that this model could generate lewd, biased or otherwise offensive outputs. Therefore, before deploying any applications of the model, developers should perform safety testing.

Caveats and Recommendations

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. Here are a couple of useful links to learn more about Intel's AI software:

Disclaimer

The license on this model does not constitute legal advice. We are not responsible for the actions of third parties who use this model. Please consult an attorney before using this model for commercial purposes.

Cite

@article{cheng2023optimize,
  title={Optimize weight rounding via signed gradient descent for the quantization of llms},
  author={Cheng, Wenhua and Zhang, Weiwei and Shen, Haihao and Cai, Yiyang and He, Xin and Lv, Kaokao and Liu, Yi},
  journal={arXiv preprint arXiv:2309.05516},
  year={2023}
}

arxiv github

This model is part of the Intel Low-Bit Open LLM Leaderboard initiative.

LeaderboardModel1/Qwen3.5-21B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking-AutoRound-W4A16-RTN

Qwen3.5-21B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-Thinking-AutoRound-W4A16-RTN

Model Details