Deepseek R1 0528 GPTQ Int4 Int8Mix Compact

🚀 DeepSeek-R1-0528-GPTQ-Int4-Int8Mix-Compact

This repository provides an Int4 + selectively-Int8 GPTQ DeepSeek-R1-0528 model, which maintains generation quality with minimal file-size overhead.

🚀 Quick Start

The base model is deepseek-ai/DeepSeek-R1-0528. This repository delivers an Int4 + selectively-Int8 GPTQ DeepSeek-R1-0528 model. Only layers that are highly sensitive to quantization remain in Int8, while the rest stay Int4, preserving generation quality with minimal file-size overhead.

Preliminary trials show that converting the entire model to pure Int4 (AWQ/GPTQ) under the quantization layout used in vLLM’s current DeepSeek-R1 implementation degrades inference accuracy and can produce faulty outputs. Layer-wise fine-grained quantization substantially mitigates this issue.

⚠️ Important Note

vLLM == 0.9.0 does not yet natively support per-layer quantization for MoE modules. We added get_moe_quant_method to gptq_marlin.py as an interim fix. Until the upstream PR is merged, please replace the original file with the one provided in this repo.

✨ Features

Variant Overview

Choose the variant that best matches your hardware and quality requirements.

Model Update Date

2025-05-31
1. fast commit

Dependencies

vllm==0.9.0
transformers==4.52.3

⚠️ Important Note

Recommend Using V0 Inference Mode: Before launching vLLM, set the environment variable

export VLLM_USE_V1=0

📦 Installation

Model Download

from huggingface_hub import snapshot_download
snapshot_download('QuantTrio/DeepSeek-R1-0528-GPTQ-Int4-Int8Mix-Compact', cache_dir="local_path")

💻 Usage Examples

Base Model Information

Paper Link

Introduction

The DeepSeek R1 model has undergone a minor version upgrade, with the current version being DeepSeek-R1-0528. In the latest update, DeepSeek R1 has significantly improved its depth of reasoning and inference capabilities by leveraging increased computational resources and introducing algorithmic optimization mechanisms during post-training. The model has demonstrated outstanding performance across various benchmark evaluations, including mathematics, programming, and general logic. Its overall performance is now approaching that of leading models, such as O3 and Gemini 2.5 Pro.

Compared to the previous version, the upgraded model shows significant improvements in handling complex reasoning tasks. For instance, in the AIME 2025 test, the model’s accuracy has increased from 70% in the previous version to 87.5% in the current version. This advancement stems from enhanced thinking depth during the reasoning process: in the AIME test set, the previous model used an average of 12K tokens per question, whereas the new version averages 23K tokens per question.

Beyond its improved reasoning capabilities, this version also offers a reduced hallucination rate, enhanced support for function calling, and better experience for vibe coding.

Evaluation Results

For all our models, the maximum generation length is set to 64K tokens. For benchmarks requiring sampling, we use a temperature of $0.6$, a top-p value of $0.95$, and generate 16 responses per query to estimate pass@1.

📚 Documentation

License

This code repository is licensed under MIT License. The use of DeepSeek-R1 models is also subject to MIT License. DeepSeek-R1 series (including Base and Chat) supports commercial use and distillation.

Citation

@misc{deepseekai2025deepseekr1incentivizingreasoningcapability,
      title={DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning}, 
      author={DeepSeek-AI},
      year={2025},
      eprint={2501.12948},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2501.12948}, 
}

Contact

If you have any questions, please raise an issue or contact us at service@deepseek.com.