As a major telecommunications operator with over 40 million mobile subscribers, SoftBank Corp. (“SoftBank”) generates large amounts of data on its network traffic and operations every day. To make its network operations more efficient and deliver services of even higher quality, SoftBank developed a Large Telecom Model (LTM). The LTM – which is trained on SoftBank’s network data, design, management and its operational know-how – substitutes tasks performed by telecommunications industry experts to increase efficiencies and reduce human error.
Main features of the LTM include:
- Knowledge and insights as SoftBank's mobile network specialist
The LTM provides knowledge and insights cultivated by SoftBank's mobile network experts, reflecting extensive network and data, along with in-depth network information annotated by in-house experts skilled in network design, management, and operation.
- 2. AI models specialized for various use cases through fine-tuning
The fine-tuning of models based on the LTM makes it possible to develop AI models specialized for various use cases in mobile network operations. Initial use cases include models developed for new base station deployments in high-density areas of Tokyo and reconfiguring existing base stations to accommodate sudden traffic surges.
- 3. Collaboration with NVIDIA
SoftBank used the NVIDIA DGX SuperPOD for distributed training during the LTM development. SoftBank will continue to collaborate with NVIDIA as it moves toward deploying the LTM.
The LTM also functions as a foundation for the “AI for RAN” initiative, which aims to enhance Radio Access Network (RAN) performance through AI. In the future, the LTM is expected to serve as a blueprint for network design and support the development of network optimization AI agents.
Going forward, SoftBank will explore ways of utilizing the LTM in its own operations, aiming to enhance its mobile network efficiency, create new services, and deliver higher-quality network experiences.
For more information, please see this press release.
(Posted on March 26, 2025)
by SoftBank News Editors
