Quant	Size (MB)	PPL	Size (%)	Accuracy (%)	PPL error rate
IQ1_S	6937	10.7622	11.10	54.42	0.15
IQ1_M	7564	9.2039	12.10	63.64	0.13
IQ2_XXS	8610	7.5917	13.78	77.15	0.11
IQ2_XS	9496	7.1233	15.19	82.23	0.10
IQ2_S	9906	7.0558	15.85	83.01	0.10
IQ2_M	10742	6.7367	17.19	86.95	0.09
Q2_K_S	10955	6.9707	17.53	84.03	0.10
Q2_K	11742	6.7073	18.79	87.33	0.09
IQ3_XXS	12244	6.3673	19.59	91.99	0.09
IQ3_XS	13070	6.2584	20.91	93.59	0.08
Q3_K_S	13725	6.2372	21.96	93.91	0.08
IQ3_S	13768	6.1756	22.03	94.85	0.08
IQ3_M	14124	6.1781	22.60	94.81	0.08
Q3_K_M	15196	6.0711	24.31	96.48	0.08
Q3_K_L	16448	6.0467	26.32	96.87	0.08
IQ4_XS	16873	5.9536	27.00	98.38	0.08
IQ4_NL	17816	5.9556	28.51	98.35	0.08
Q4_0	17844	5.9980	28.55	97.65	0.08
Q4_K_S	17914	5.9378	28.66	98.64	0.08
Q4_K_M	18931	5.9273	30.29	98.82	0.08
Q4_1	19683	5.9349	31.49	98.69	0.08
Q5_K_S	21589	5.8895	34.54	99.45	0.08
Q5_0	21657	5.9057	34.65	99.18	0.08
Q5_K_M	22184	5.8872	35.49	99.49	0.08
Q5_1	23495	5.8896	37.59	99.45	0.08
Q6_K	25640	5.8667	41.02	99.84	0.08
Q8_0	33207	5.8646	53.13	99.88	0.08
F16	62499	5.8574	100	100	0.08

QwQ-32B-Preview

Introduction

QwQ-32B-Preview is an experimental research model developed by the Qwen Team, focused on advancing AI reasoning capabilities. As a preview release, it demonstrates promising analytical abilities while having several important limitations:

Language Mixing and Code-Switching: The model may mix languages or switch between them unexpectedly, affecting response clarity.
Recursive Reasoning Loops: The model may enter circular reasoning patterns, leading to lengthy responses without a conclusive answer.
Safety and Ethical Considerations: The model requires enhanced safety measures to ensure reliable and secure performance, and users should exercise caution when deploying it.
Performance and Benchmark Limitations: The model excels in math and coding but has room for improvement in other areas, such as common sense reasoning and nuanced language understanding.

Specification:

Type: Causal Language Models
Training Stage: Pretraining & Post-training
Architecture: transformers with RoPE, SwiGLU, RMSNorm, and Attention QKV bias
Number of Parameters: 32.5B
Number of Paramaters (Non-Embedding): 31.0B
Number of Layers: 64
Number of Attention Heads (GQA): 40 for Q and 8 for KV
Context Length: Full 32,768 tokens

For more details, please refer to our blog. You can also check Qwen2.5 GitHub, and Documentation.

Requirements

The code of Qwen2.5 has been in the latest Hugging face transformers and we advise you to use the latest version of transformers.

With transformers<4.37.0, you will encounter the following error:

KeyError: 'qwen2'

Quickstart

Here provides a code snippet with apply_chat_template to show you how to load the tokenizer and model and how to generate contents.

python

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "Qwen/QwQ-32B-Preview"

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "How many r in strawberry."
messages = [
    {"role": "system", "content": "You are a helpful and harmless assistant. You are Qwen developed by Alibaba. You should think step-by-step."},
    {"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)

generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=512
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]

Citation

If you find our work helpful, feel free to give us a cite.

@misc{qwq-32b-preview,
    title = {QwQ: Reflect Deeply on the Boundaries of the Unknown},
    url = {https://qwenlm.github.io/blog/qwq-32b-preview/},
    author = {Qwen Team},
    month = {November},
    year = {2024}
}

@article{qwen2,
      title={Qwen2 Technical Report}, 
      author={An Yang and Baosong Yang and Binyuan Hui and Bo Zheng and Bowen Yu and Chang Zhou and Chengpeng Li and Chengyuan Li and Dayiheng Liu and Fei Huang and Guanting Dong and Haoran Wei and Huan Lin and Jialong Tang and Jialin Wang and Jian Yang and Jianhong Tu and Jianwei Zhang and Jianxin Ma and Jin Xu and Jingren Zhou and Jinze Bai and Jinzheng He and Junyang Lin and Kai Dang and Keming Lu and Keqin Chen and Kexin Yang and Mei Li and Mingfeng Xue and Na Ni and Pei Zhang and Peng Wang and Ru Peng and Rui Men and Ruize Gao and Runji Lin and Shijie Wang and Shuai Bai and Sinan Tan and Tianhang Zhu and Tianhao Li and Tianyu Liu and Wenbin Ge and Xiaodong Deng and Xiaohuan Zhou and Xingzhang Ren and Xinyu Zhang and Xipin Wei and Xuancheng Ren and Yang Fan and Yang Yao and Yichang Zhang and Yu Wan and Yunfei Chu and Yuqiong Liu and Zeyu Cui and Zhenru Zhang and Zhihao Fan},
      journal={arXiv preprint arXiv:2407.10671},
      year={2024}
}

ThomasBaruzier/QwQ-32B-Preview-GGUF

Llama.cpp imatrix quantizations of Qwen/QwQ-32B-Preview

Perplexity table (the lower the better)

QwQ-32B-Preview

Introduction

Requirements

Quickstart

Citation

ThomasBaruzier/
QwQ-32B-Preview-GGUF