A Study on the Design of MSP Systems for Multiple LLM Integration Services
Hae-Jun Lee · Asia-pacific Journal of Convergent Research Interchange · 2025
The need for resource efficiency to reduce the demand cost of service users is increasing due to the recent multiplexing of Generative AI platforms.With the emergence of various LLM (Large Language Model) services such as ChatGPT, Claude, and Gemini, companies increasingly demand to select and use the optimal LLM for their respective business purposes.However, when multiple LLMs are used simultaneously, integrated management and operation are difficult due to different API (Application Programmable Interface) specifications, billing systems, and operation policies for each service.In this study, by developing an MSP (Managed Service Provider) system for the integrated operation of multiple LLM services, companies' service utilization efficiency can be improved by 50% or more, and operating costs can be reduced by more than 30%.In particular, although the monthly subscription cost of major LLM services is currently $20-25, integrated subscription management through MSP can lead to securing purchasing power through economies of scale and maximize cost reduction.Furthermore, investing the resources secured through these cost savings in the development of its own SLM (small language model) can lead to securing competitiveness in AI models unique to companies in the mid-to-long-term.