Author(s)

HaiLong Liao

Affiliation(s)

School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing 100049, China.

Corresponding Author

HaiLong Liao

ABSTRACT

This paper deeply analyzes the DeepSeek large - scale model, comprehensively elaborating on its technical architecture, training mechanism, performance, application fields, as well as the challenges it faces and future development directions. Through the research on the DeepSeek series of models, it reveals their innovations and important values in the field of artificial intelligence. The research shows that the DeepSeek large - scale model, with its unique technical advantages, demonstrates excellent performance in tasks such as natural language processing, code generation, and multimodal understanding. It provides new ideas and methods for promoting the development and application of artificial intelligence technology.

KEYWORDS

DeepSeek; Large language model; Artificial intelligence; Multimodality

CITE THIS PAPER

HaiLong Liao. DeepSeek large - scale model: technical analysis and development prospect. Journal of Computer Science and Electrical Engineering. 2025, 7(1): 33-37. DOI: https://doi.org/10.61784/jcsee3035.

REFERENCES

[1] Rohan Paul. DeepSeek - V3's Architectural Revolution: Rewriting the Economics of Large Language Model Training. 2024.  Retrieved from https://rohanpaul.substack.com/p/deepseek-v3-technical-report-they

[2] Vaswani, A, Shazeer, N, Parmar, N, et al. Attention Is All You Need. Advances in Neural Information Processing Systems, 2017.

[3] DeepSeek-V3 Technical Report. It is authored by DeepSeek-AI, 2024. DOI: https://doi.org/10.48550/arXiv.2412.19437. Retrieved from https://arxiv.org/abs/2412.19437v1. Project homepage: https://github.com/deepseek-ai/DeepSeek-V3.

[4] Elmo. DeepSeek - VL: New Open Source Vision - Language Models! Medium (Medium Reviews). 2024. Retrieved from https://medium.com/@elmo92/deepseek-vl-new-open-source-vision-language-models-32bc77fa4647

[5] Xiaokang Chen, Zhiyu Wu, Xingchao Liu, et al. Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling, 2025. Retrieved from https://arxiv.org/abs/2501.17811v1

[6] DeepSeek-AI, Daya Guo, Dejian Yang, et al. DeepSeek - R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning. 2025. DOI: https://doi.org/10.48550/arXiv.2501.12948. Retrieved from https://arxiv.org/pdf/2501.12948

[7] DeepSeek-AI, Xiao Bi, Deli Chen, et al. DeepSeek LLM: Scaling Open-Source Language Models with Longtermism. 2024. Retrieved from https://arxiv.org/abs/2401.02954