gemma-4-26B-A4B-AutoRound-MXFP8-ModelFree

Model Details

This model is a MXFP8 quantization of google/gemma-4-26B-A4B generated by agent_optimize. Please follow the license of the original model.

Quantization Details

Attribute	Value
Base Model	google/gemma-4-26B-A4B
Quantization Tool	agent_optimize
Quantization Scheme	MXFP8

Evaluation Results

Task	Accuracy
gsm8k	0.7346
hellaswag	0.6361
mmlu	0.7446
mmlu_abstract_algebra	0.4300
mmlu_anatomy	0.7852
mmlu_astronomy	0.8421
mmlu_business_ethics	0.7600
mmlu_clinical_knowledge	0.8415
mmlu_college_biology	0.8889
mmlu_college_chemistry	0.5700
mmlu_college_computer_science	0.5900
mmlu_college_mathematics	0.4000
mmlu_college_medicine	0.7283
mmlu_college_physics	0.5490
mmlu_computer_security	0.7300
mmlu_conceptual_physics	0.8128
mmlu_econometrics	0.6491
mmlu_electrical_engineering	0.7172
mmlu_elementary_mathematics	0.6349
mmlu_formal_logic	0.5635
mmlu_global_facts	0.4700
mmlu_high_school_biology	0.9097
mmlu_high_school_chemistry	0.7192
mmlu_high_school_computer_science	0.8600
mmlu_high_school_european_history	0.8424
mmlu_high_school_geography	0.9444
mmlu_high_school_government_and_politics	0.9430
mmlu_high_school_macroeconomics	0.8282
mmlu_high_school_mathematics	0.4222
mmlu_high_school_microeconomics	0.8866
mmlu_high_school_physics	0.5629
mmlu_high_school_psychology	0.9321
mmlu_high_school_statistics	0.7269
mmlu_high_school_us_history	0.9069
mmlu_high_school_world_history	0.8861
mmlu_human_aging	0.7713
mmlu_human_sexuality	0.8626
mmlu_humanities	0.6616
mmlu_international_law	0.8760
mmlu_jurisprudence	0.8796
mmlu_logical_fallacies	0.7485
mmlu_machine_learning	0.6429
mmlu_management	0.9320
mmlu_marketing	0.9274
mmlu_medical_genetics	0.8800
mmlu_miscellaneous	0.9029
mmlu_moral_disputes	0.8237
mmlu_moral_scenarios	0.2972
mmlu_nutrition	0.8333
mmlu_other	0.8091
mmlu_philosophy	0.8424
mmlu_prehistory	0.8735
mmlu_professional_accounting	0.6348
mmlu_professional_law	0.6115
mmlu_professional_medicine	0.8640
mmlu_professional_psychology	0.8317
mmlu_public_relations	0.7727
mmlu_security_studies	0.8367
mmlu_social_sciences	0.8651
mmlu_sociology	0.8657
mmlu_stem	0.6873
mmlu_us_foreign_policy	0.9100
mmlu_virology	0.5602
mmlu_world_religions	0.8830
piqa	0.8232

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 = "gemma-4-26B-A4B-AutoRound-MXFP8-ModelFree"

# 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 gemma-4-26B-A4B-AutoRound-MXFP8-ModelFree \
    --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:

• Intel Neural Compressor
• AutoRound

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.