Qwen3.6-27B-AutoRound-MXFP4-Tuning

Model Details

This model is a MXFP4 (Microscaling FP4) quantization of Qwen/Qwen3.6-27B generated by TUNING. Please follow the license of the original model.

Quantization Details

Attribute	Value
Base Model	Qwen/Qwen3.6-27B
Quantization Tool	TUNING
Quantization Scheme	MXFP4
Quantized Size	21231 MB

Evaluation Results

Task	Accuracy
hellaswag	0.6201
mmlu	0.8347
mmlu_abstract_algebra	0.7100
mmlu_anatomy	0.8074
mmlu_astronomy	0.9211
mmlu_business_ethics	0.8000
mmlu_clinical_knowledge	0.9019
mmlu_college_biology	0.9514
mmlu_college_chemistry	0.6800
mmlu_college_computer_science	0.8000
mmlu_college_mathematics	0.7100
mmlu_college_medicine	0.8439
mmlu_college_physics	0.6961
mmlu_computer_security	0.8300
mmlu_conceptual_physics	0.9064
mmlu_econometrics	0.8070
mmlu_electrical_engineering	0.8345
mmlu_elementary_mathematics	0.8810
mmlu_formal_logic	0.7698
mmlu_global_facts	0.5900
mmlu_high_school_biology	0.9355
mmlu_high_school_chemistry	0.8177
mmlu_high_school_computer_science	0.9100
mmlu_high_school_european_history	0.8545
mmlu_high_school_geography	0.9495
mmlu_high_school_government_and_politics	0.9637
mmlu_high_school_macroeconomics	0.9128
mmlu_high_school_mathematics	0.6630
mmlu_high_school_microeconomics	0.9496
mmlu_high_school_physics	0.7815
mmlu_high_school_psychology	0.9468
mmlu_high_school_statistics	0.8333
mmlu_high_school_us_history	0.9167
mmlu_high_school_world_history	0.9409
mmlu_human_aging	0.8206
mmlu_human_sexuality	0.9084
mmlu_humanities	0.7809
mmlu_international_law	0.9504
mmlu_jurisprudence	0.8796
mmlu_logical_fallacies	0.9141
mmlu_machine_learning	0.7321
mmlu_management	0.8835
mmlu_marketing	0.8846
mmlu_medical_genetics	0.9500
mmlu_miscellaneous	0.9349
mmlu_moral_disputes	0.7688
mmlu_moral_scenarios	0.7061
mmlu_nutrition	0.9020
mmlu_other	0.8597
mmlu_philosophy	0.8103
mmlu_prehistory	0.8858
mmlu_professional_accounting	0.7660
mmlu_professional_law	0.7027
mmlu_professional_medicine	0.9449
mmlu_professional_psychology	0.8644
mmlu_public_relations	0.7909
mmlu_security_studies	0.7959
mmlu_social_sciences	0.9012
mmlu_sociology	0.9204
mmlu_stem	0.8256
mmlu_us_foreign_policy	0.9400
mmlu_virology	0.5422
mmlu_world_religions	0.8889
piqa	0.8063

How to Use

HF Usage

Step 1: Install AutoRound

pip install auto-round

Step 2: Load and run the quantized model

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "Qwen3.6-27B-AutoRound-MXFP4-Tuning"

# 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

vllm serve Qwen3.6-27B-AutoRound-MXFP4-Tuning \
    --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.