Large Telecom Model for AI-Native Networks Demoed at MWC
At MWC Barcelona, a collaboration including Northeastern University and SoftBank Corp. demonstrated an agentic, AI-native Radio Access Network (AI-RAN). The system is powered by a 'Large Telecom Model' (LTM), a specialized foundation model designed for managing wireless networks, signaling a move toward domain-specific agentic AI.
- Agentic AI in telecommunications introduces autonomous AI agents that can understand intent, reason across complex systems, and act independently to manage network operations. This differs from earlier AI, which was limited to predictions or recommendations, by allowing the system to execute multi-step workflows and learn from the results. - The Large Telecom Model (LTM) is a type of domain-specific foundation model, meaning it's an AI model trained on a massive, specialized dataset specifically for telecommunications. This allows it to understand the unique language, protocols, and patterns of wireless networks far better than a general-purpose AI model. - An AI-Radio Access Network (AI-RAN) integrates AI into every layer of the network's hardware and software, from the physical radio processing to resource management. This allows the network to perform real-time optimizations for things like spectrum efficiency, traffic steering, and energy consumption. - The demonstration at MWC Barcelona 2026 is a collaboration between Northeastern University's Institute for Intelligent Networked Systems, SoftBank Corp., Keysight Technologies, and Ztouch Networks. Northeastern has been a key player in open-source wireless research, leading initiatives like Open6G+AI to accelerate innovation in the field. - SoftBank is a founding member of the AI-RAN Alliance and is heavily invested in defining the architecture for AI-native networks. The company has been collaborating with partners like Samsung and Ericsson on AI-RAN technologies with the goal of creating a new social infrastructure where AI and humans coexist. - A key challenge this technology addresses is the complexity of managing concurrent AI and RAN workloads on the same infrastructure. The collaboration aims to validate solutions that can dynamically allocate resources between these different tasks, ensuring reliability and performance for both network functions and new AI-driven services at the network edge. - The move toward AI-native networks is seen as a critical step for the rollout of 6G and the support of "physical AI" applications like autonomous vehicles and robotics, which require distributed computing with extremely low latency and high reliability. - Other major telecom players are also heavily invested in AI-RAN, with companies like SK Telecom, NTT DOCOMO, and Ericsson showcasing their own AI-native solutions and research at MWC 2026. The overall theme of the conference is "The IQ Era," reflecting the industry-wide shift toward smarter, more automated networks and devices.
