Khalifa University launches RF-GPT, a world-first radio-frequency AI model

The Brew News, your leading alt-news source for the latest updates in sports, entertainment, business, tech and IPL straight up from Dubai, UAE. As the leading news portal in the Middle East, we're committed to bringing you latest insights from across the GCC, South Asia and the World.

Khalifa University’s Digital Future Institute has unveiled RF-GPT, a groundbreaking radio-frequency AI language model that can read and interpret wireless signals in plain language, marking a major leap forward for spectrum intelligence and next-generation 6G networks.

ABU DHABI: A team of researchers at Khalifa University of Science and Technology has built something the telecommunications world has never seen before. RF-GPT is a radio-frequency AI language model that can look at wireless signals and answer questions about them in plain language, closing a gap that has long held back AI in the telecom space.

Traditional language models work with text and structured data. RF-GPT works differently. It converts radio signals into visual patterns, then analyses those patterns to make sense of what is happening across the wireless spectrum. The radio-frequency AI language model performed consistently well across a range of real-world tasks, outperforming existing models by as much as 75.4 per cent on spectrogram analysis. It also correctly counted the number of signals in a spectrogram nearly 98 per cent of the time, a result that general-purpose AI tools rarely come close to achieving.

The model was trained on roughly 625,000 computer-generated radio signal examples and is built to support telecoms operators, network engineers, and spectrum regulators. It can identify signal types, detect overlapping transmissions, recognise wireless standards, estimate device activity in Wi-Fi networks, and extract data from 5G signals.

Professor Merouane Debbah, who led the project, described RF-GPT as a turning point for spectrum intelligence, one that moves the field away from isolated, task-specific pipelines and toward a unified interface where the physical layer can be queried in natural language. He said the model opens the door to AI-native radio systems capable of supporting network optimisation and policy decisions, and represents a crucial step toward future 6G infrastructure.

The work directly supports the UAE National Artificial Intelligence Strategy and contributes to the country’s broader goals in digital infrastructure and human capital development.

For in-depth analysis, exclusive stories and comprehensive coverage of the latest happenings in the UAE and beyond, visit our website at www.thebrewnews.com. We value your readership and look forward to bringing you more breaking news and stories.