SoftBank Corp. has introduced a Large Telecom Model (LTM) designed for the telecom industry, utilizing its extensive datasets from network operations. This generative Artificial Intelligence (AI) model enhances inference capabilities related to the design and management of cellular networks. SoftBank plans to integrate the LTM into its operations while conducting further research and development.

By fine-tuning the LTM, SoftBank has created specialized models that optimize base station configurations, achieving over 90% accuracy in predictions verified by in-house experts. This method significantly reduces the time required for operational changes, from days to minutes, minimizing human error and operational costs.

SoftBank's LTM serves as a foundation for the “AI for RAN” initiative, which aims to enhance Radio Access Network (RAN) performance. The model's design was optimized using NVIDIA’s technologies, providing notable improvements in performance metrics.

Ryuji Wakikawa, Vice President at SoftBank, highlighted the model's potential to transform telecommunications operations, enabling specialized AI models for various processes and improving overall network performance. Chris Penrose of NVIDIA noted the model's capability to streamline network operations for telecom providers globally.

SoftBank aims to deploy the LTM to improve efficiency, develop new services, and enhance network quality while continuing collaborations with partners to advance next-generation network capabilities.