As large language models are increasingly applied in various places, such as centralised cloud services, terminals, edge servers and private industry networks, issues such as model size, GPU memory capacity, memory bandwidth, first-word latency and energy consumption constraints have become progressively more serious. This paper studies the light-weight deployment and inference acceleration of large language models in resource-constrained environments. It summarizes the research progress in the past three years for low-bit quantization, lightweight model design, key-value cache compression, attention operator optimisation and parallel decoding, and builds a deployment framework of "model compression - cache governance - operator fusion - scheduling coordination". Therefore, it has been concluded that a single compression strategy is unable to meet all the demands for accuracy, low latency and energy efficiency. For practical implementation at the edge, a verifiable, rollbackable and portable inference optimisation process needs to be established, and the joint objectives of task quality degradation, peak memory usage, throughput and energy consumption should be optimised.

Yiling Yuan, Research on Lightweight Deployment and Reasoning Acceleration of Large Language Models in Resource Constrained Environments. International Journal of Social Sciences and Economic Management (2026), Vol. 7, Issue 2: 78-87. https://doi.doi.org/10.3807/IJSSEM.2026.070208.

[1] Lin J., Tang J., Tang H., Yang S., Chen WM, Wang WC, Xiao G., Dang X., Gan C., Han S. AWQ: Activation-aware Weight Quantization for On-Device LLM Compression and Acceleration. Proceedings of Machine Learning and Systems, 2024, 6.

[2] Liu Z., Zhao C., Iandola F., Lai C., Tian Y., Fedorov I., Xiong Y., Chang E., Shi Y., Krishnamoorthi R., Lai L., Chandra V. MobileLLM: Optimizing Sub-billion Parameter Language Models for On-Device Use Cases. Proceedings of the 41st International Conference on Machine Learning, 2024.

[3] Kwon W., Li Z., Zhuang S., Sheng Y., Zheng L., Yu CH, Gonzalez JE, Zhang H., Stoica I. Efficient Memory Management for Large Language Model Serving with PagedAttention. Proceedings of the ACM SIGOPS 29th Symposium on Operating Systems Principles, 2023.

[4] Liu Z., Yuan J., Jin H., Zhong S., Xu Z., Braverman V., Chen B., Hu X. KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache. Proceedings of the 41st International Conference on Machine Learning, 2024.

[5] Zhang, Y. (2026). Exploration of Enterprise Big Data Microservice Architecture Based on Domain-Driven Design (DDD). Procedia Computer Science, 282, 1994-2003.

[6] Liu, X. (2026). Research on the Application of Artificial Intelligence in Consumer Privacy Data Protection. Procedia Computer Science, 279, 956-965.

[7] Jing, X. (2025, November). In-Depth Analysis of the Optimization Decision Model of Financial Investment Portfolio Based on Machine Learning Algorithms in Python Scikit-Learn Library. In International Conference on Computational Technologies for Research in Data Analytics (pp. 130-140). Cham: Springer Nature Switzerland.

[8] Yu, X. (2026). Exploration of Multi-Channel Conversion Path Optimization Methods Based on A/B Testing.

[9] Han, W. (2025, November). An Empirical Study on Building a High-precision Forecasting System for Financial Market Trends Based on the PyTorch Deep Learning Framework. In International Conference on Computational Technologies for Research in Data Analytics (pp. 518-528). Cham: Springer Nature Switzerland.

[10] Yuan, Y. (2026). Research on Memory Management and Dynamic Reasoning Methods for Intelligent Agents Based on Large Language Models.