RF-GPT turns radio signals into visual patterns for AI to interpret, moving spectrum intelligence from isolated, task-specific RF pipelines toward a unified RF-language interface.
Khalifa University of Science and Technology’s Digital Future Institute is launching RF-GPT, an AI language model that can interpret wireless signals. The Institute says the model the first of its kind in the world which overcomes a major limitation of AI in telecom: language models typically operate only on text and structured network data.
RF-GPT works by turning radio signals into visual patterns that AI can understand. Once converted, AI systems can analyze those patterns and answer questions about what is happening in the wireless spectrum using plain language.
The foundation model directly contributes to the United Arab Emirates’ (UAE’s) National Artificial Intelligence Strategy, laying the groundwork for more autonomous and intelligent wireless networks.
In trials the model showed consistent improvements in performance in radio frequency spectrogram tasks, outperforming baseline models by up to 75.4% and demonstrating “a strong understanding” of radio frequency according to the Institute.
Strong performance
The model performed strongly across tasks such as identifying signal types, detecting overlapping transmissions, recognising wireless standards, estimating device usage in Wi-Fi networks, and extracting data from 5G signals.
RF-GPT also correctly counted the number of signals in a spectrogram ~98% of the time, which general-purpose AI models almost never achieve.
The model was trained using about 625,000 examples of computer-generated radio signals. It is designed for network operators, network engineering teams and spectrum authorities, all of whom support increasingly complex wireless environments.
The project was developed by researchers at Khalifa University, led by Professor Merouane Debbah, Senior Director, Digital Future Institute. The team includes Post Doctoral Fellows Hang Zou, Yu Tian, Research Scientists Dr Lina Bariah, Khalifa University, Dr. Samson Lasaulce, L’Université de Lorraine, and Dr Chongwen Huang and PhD student Bohao Wang, both from Zhejiang University
Imagine what it will do next
Professor Merouane Debbah said, “RF-GPT represents a turning point for spectrum intelligence, moving from isolated, task-specific radio frequency pipelines toward a unified RF-language interface. We gave a language model its first glimpse of the electromagnetic spectrum and the view is already remarkable. Imagine what it will see next.
“By making the physical layer quarriable in natural language, we open the door to AI-native radio systems where RF perception can directly support network optimisation and policy decisions, a crucial step toward future AI-native 6G networks.”
