s
starmountain/Qwen2.5-0.5B
Model Introduction
Files & versions
Pull Requests
Discussion
Analysis

Qwen2.5-0.5B

Introduction

Qwen2.5 is the latest series of Qwen large language models. For Qwen2.5, we release a number of base language models and instruction-tuned language models ranging from 0.5 to 72 billion parameters. Qwen2.5 brings the following improvements upon Qwen2:

  • Significantly more knowledge and has greatly improved capabilities in coding and mathematics, thanks to our specialized expert models in these domains.
  • Significant improvements in instruction following, generating long texts (over 8K tokens), understanding structured data (e.g, tables), and generating structured outputs especially JSON. More resilient to the diversity of system prompts, enhancing role-play implementation and condition-setting for chatbots.
  • Long-context Support up to 128K tokens and can generate up to 8K tokens.
  • Multilingual support for over 29 languages, including Chinese, English, French, Spanish, Portuguese, German, Italian, Russian, Japanese, Korean, Vietnamese, Thai, Arabic, and more.

This repo contains the base 0.5B Qwen2.5 model, which has the following features:

  • Type: Causal Language Models
  • Training Stage: Pretraining
  • Architecture: transformers with RoPE, SwiGLU, RMSNorm, Attention QKV bias and tied word embeddings
  • Number of Parameters: 0.49B
  • Number of Paramaters (Non-Embedding): 0.36B
  • Number of Layers: 24
  • Number of Attention Heads (GQA): 14 for Q and 2 for KV
  • Context Length: Full 32,768 tokens

We do not recommend using base language models for conversations. Instead, you can apply post-training, e.g., SFT, RLHF, continued pretraining, etc., on this model.

For more details, please refer to our blog, 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'

Evaluation & Performance

Detailed evaluation results are reported in this blog.

For requirements on GPU memory and the respective throughput, see results here.


Ascend NPU Adaptation Report

Verification Environment: Huawei Ascend NPU (Atlas 800 A2)
Framework: vLLM-Ascend 0.18.0rc1
Verification Date: 2026-05-13
Status: All tests passed

Model Configuration

PropertyValue
Model NameQwen2.5-0.5B
ArchitectureQwen2ForCausalLM
Parameters0.5B
Layers24
Max Model Length32,768

Environment Check

CheckStatusDetails
NPU SMIPassed2x Ascend910, Health OK, 64GB HBM each
vLLM InstalledPassedvLLM 0.18.0+empty
vLLM-AscendPassed0.18.0rc1
torch_npuPassedAvailable

Deployment Configuration

ParameterValue
Tensor Parallel Size1
GPU Memory Utilization0.85
Compilation ModeFULL_DECODE_ONLY
Async Schedulingtrue
Served Model Nameqwen2.5-0.5b

Functional Verification

EndpointStatusDetails
/v1/modelsSuccessModel listed correctly with max_model_len=32768
/v1/completionsSuccessprompt_tokens=6, completion_tokens=16
/v1/chat/completionsSuccessprompt_tokens=29, completion_tokens=50

Performance Benchmarks

1. Serve Benchmark (Online Serving)

MetricValue
Request Rate5.0 req/s
Total Prompts100
Successful / Failed100 / 0
Duration20.90 s
Request Throughput4.78 req/s
Output Token Throughput956.86 tok/s
Peak Output Token Throughput2,006.00 tok/s
Total Token Throughput1,913.71 tok/s
Peak Concurrent Requests17

Latency Breakdown:

MetricMeanMedianP99
TTFT (ms)42.6339.7561.70
TPOT (ms)5.075.065.97
ITL (ms)5.074.6224.26

2. Latency Benchmark (Single Request)

MetricValue
Input Length200 tokens
Output Length200 tokens
Batch Size1
Iterations10
Avg Latency2.299 s
P50 Latency2.291 s
P99 Latency2.471 s
Engine Init Time17.49 s
Graph Capture Time4.0 s
KV Cache Size4,442,752 tokens
Max Concurrency (32K)135.58

3. Throughput Benchmark (Offline Batch)

MetricValue
Input Length200 tokens
Output Length200 tokens
Num Prompts64
Requests/s25.40
Total Tokens/s29,259.35
Output Tokens/s3,251.04
Total Input Tokens65,536
Total Output Tokens8,192
Engine Init Time17.17 s
Graph Capture Time4.0 s

Verification Log

  1. Environment check passed (2x Ascend910, vllm-ascend 0.18.0rc1)
  2. Model downloaded from ModelScope to local cache
  3. vLLM API server deployed on NPU 0 (PID 1006)
  4. Functional verification passed (models, completions, chat.completions)
  5. Serve benchmark passed (100/100 requests successful, 4.78 req/s)
  6. Latency benchmark passed on NPU 1 (P50=2.29s, P99=2.47s)
  7. Throughput benchmark passed on NPU 1 (25.40 req/s, 3251 output tok/s)
  8. Resource cleanup completed, no residual processes

Citation

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

@misc{qwen2.5,
    title = {Qwen2.5: A Party of Foundation Models},
    url = {https://qwenlm.github.io/blog/qwen2.5/},
    author = {Qwen Team},
    month = {September},
    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}
}