Global Mobile Industry Launches Telco AI Platform

The "Open Telco AI" platform was created to address the significant performance gap of general-purpose AI models in the telecommunications sector. These advanced AIs often struggle to interpret complex network data, understand dense technical standards, and automate network operations with the necessary accuracy. This has limited progress, with only 16% of generative AI deployments in the telecom industry being applied to network operations. Founding supporters of the initiative include AT&T and AMD, who are making significant early contributions. AT&T is releasing a family of open-source, telco-grade models trained on publicly available data, which are designed to be hardware and cloud-agnostic. AMD, in partnership with its cloud partner TensorWave, is providing the necessary computing power for model training, fine-tuning, and evaluation through its GPU platforms. The initiative is further supported by a wide range of participants, including major global mobile operators like China Mobile, Deutsche Telekom, Telefónica, and Vodafone. Tech giants such as Google Cloud, IBM, and Nvidia are also contributing, alongside numerous academic institutions that will provide valuable research and datasets. A key component of the platform is the "Telco Capability Index," a new benchmark to measure the performance of AI models on an expanding set of telecom-specific tasks. This will allow for a clear, standardized evaluation of which AI systems are truly "telco-grade." A leaderboard will publicly track model performance against these industry-specific benchmarks. The ultimate goal is to create more autonomous networks that can predict and prevent issues, optimize performance in real-time, and reduce operational friction. For consumers, this translates to more reliable connectivity, faster rollout of new technologies like 5G and 6G, and improved customer service experiences. The platform will provide access to a library of curated, telecom-specific data for training AI, including knowledge graphs and materials from standards bodies. It will also feature specialized models beyond AT&T's contribution, such as a radio-frequency language model from Khalifa University called RFGPT and a Large Telco Model from AdaptKey AI.