Open Telco AI Leaderboard Release 3 launched at MWC Barcelona
GSMA’s telecom AI leaderboard expands with new operator tasks and models
#UAE #LLMs – Global telecom association GSMA has released the third iteration of the Open Telco AI Leaderboard, an open benchmark platform evaluating and ranking large language models (LLMs) for telecom applications. First launched in November 2024 by GSMA, Khalifa University and Huawei Paris Research Centre, the leaderboard has expanded with Release 3 to include new telecom tasks reflecting real operator workflows and a broader mix of open and proprietary models. The announcement was made at MWC Barcelona, which took place this week, advancing the goal of building open, industry-grade AI foundations for the global telecommunications sector.
SO WHAT? – Telecommunications is one of the most technically complex domains for AI model development, encompassing network operations, configuration management, customer service and knowledge management across vast and highly specialised subject matter. The absence of widely accepted evaluation benchmarks has been a significant obstacle to progress, making it difficult for developers and operators alike to compare and validate telecom-specific LLMs effectively. The Open Telco AI Leaderboard is filling that gap, and with each release it is becoming a more credible and comprehensive standard for the industry.
Global telecom association GSMA has released the third iteration of the Open Telco AI Leaderboard, an open benchmark platform evaluating and ranking large language models (LLMs) for telecom applications. The Digital Future Institute (DFI) of Abu Dhabi-based research university Khalifa University was a key contributor to the project.
GSMA, Khalifa University and Huawei Paris Research Centre launched the first Open Telco AI Leaderboard in November 2024 during the 6G Summit Abu Dhabi.
Release 3 of the leaderboard expands the benchmark with new telecom tasks designed to reflect real operator workflows, alongside a broader mix of open and proprietary models, enabling more realistic and practically relevant evaluation of AI in live telecom environments.
The leaderboard evaluates telecom LLMs across key attributes including the ability to answer telecom-related queries with precision, mathematical reasoning capability, network operations performance, and energy efficiency: criteria directly aligned with real operator requirements.
The platform supports evaluation of methods including fine-tuning and retrieval-augmented generation (RAG) for targeted telecom applications, such as customer service chatbots, network configuration tools and knowledge management.
The leaderboard is a key component of GSMA’s broader OpenTelco AI initiative, which was just launched at MWC Barcelona this week. The initiative aims to build open foundations for telco-grade AI spanning models, datasets, benchmarks and collaborative development across the global industry.
The Digital Future Institute has contributed telecom-specific AI models to the ecosystem, including RF-GPT, a model focused on radio frequency applications, and TelecomGPT-R1, alongside domain datasets and evaluation tasks that support the broader leaderboard framework.
The Open Telco AI Leaderboard addresses a long-standing gap in the telecom AI sector, which has lacked widely accepted evaluation benchmarks enabling developers and operators to compare telecom-specific LLMs effectively.
ZOOM OUT – The release of the Open Telco AI Leaderboard’s third iteration formed part of the GSMA’s launch of the Open Telco AI initiative on March 2nd at MWC Barcelona. The fully global telecom industry initiative includes a new portal for telco open models, data, compute and tools. The scale of industry support is considerable, with founding partners including US telecoms giant AT&T, which is releasing a family of open telco-specific AI models, and semiconductor company AMD, providing compute capacity for model training and inference through its GPU platforms and cloud partner TensorWave. Khalifa University’s Digital Future Institute (DFI) is a key contributor to the initiative, leading the Network Management and Configuration Group within the Open Telco AI programme. The GSMA aims to establish Open Telco AI as the definitive open collaboration platform for telco-grade AI development.
[Written and edited with the assistance of AI]
LINKS
Open Telco AI Leaderboard (Hugging Face)
Open Telco AI portal (GSMA site)
Read more about Khalifa University’s work on telecom AI:
Khalifa University unveils breakthrough RF AI model (Middle East AI News)
UAE University, KU release first open 6G AI benchmark (Middle East AI News)
GSMA, Khalifa University to update TelecomGPT (Middle East AI News)
Khalifa University announces telecom AI benchmarks (Middle East AI News)
New telecom LLM leaderboard project (Middle East AI News)