GSMA, Khalifa University to develop new TelecomGPT version
TelecomGPT chat interface and Open Telco Knowledge Graph to be released online
#Qatar #UAE #LLMs - GSMA Foundry, the innovation arm of the global mobile industry organisation GSMA, and Abu Dhabi research university Khalifa University of Science and Technology have announced a strategic collaboration to develop a new version of TelecomGPT, a telecom-focused large language model, alongside specialised AI data assets and benchmarking frameworks for the mobile sector. The first phase of the partnership, will see Khalifa University’s 6G Research Centre release open telco assets including an advanced LLM with chat interface hosted on LightOn and an Open Telco Knowledge Graph focused on 3GPP documentation. TelecomGPT addresses critical gaps in general-purpose AI models, which GSMA benchmarks show struggle with telecom-specific challenges including standards interpretation and network troubleshooting. The news was announced on the opening day of MWC Doha in Qatar.
SO WHAT? - The GSMA Foundry and Khalifa University announced the first version of TelecomGPT last year, which was shared with GSMA members and research partners for review. A new set of telecom industry benchmarks were announced this year to allow methodical evaluation of telecom AI modes. The new partnership between the GSMA and the university builds on this early work and sets strategic goals for future research. As an early deliverable, an online chat version of TelecomGPT, plus an Open Telco Knowledge Graph are being released to provide access across more telecom and related industry audiences.
Here are some key points related to the partnership announcement:
GSMA Foundry and Khalifa University announced a strategic collaboration to develop a new version of TelecomGPT, a telecommunications-focused large language model. The announcement was made on the opening day of MWC Doha.
The partnership will release TelecomGPT chat, an advanced telco-first large language model, as a interface hosted on LightOn, alongside an Open Telco Knowledge Graph focused on 3GPP documentation built using LightOn compute and hosted on Hugging Face.
The GSMA’s Open-Telco LLM Benchmarks demonstrate that even advanced general-purpose large language models struggle with telecom-specific tasks including interpreting industry standards, network troubleshooting and handling technical telecommunications knowledge, creating a performance gap the collaboration aims to address.
Khalifa University’s 6G Research Centre will provide infrastructure and expertise to develop AI solutions in collaboration with GSMA, combining academic research capabilities with industry requirements to create robust, safe and energy-efficient models.
The GSMA Open-Telco LLM Benchmarks initiative, launched at MWC Barcelona 2025, has evolved into a collaborative platform for leading mobile network operators to experiment with and validate LLMs tailored to telecom applications, featuring a leaderboard tracking model performance.
The next phase of the benchmarking community will expand to include holistic evaluation metrics such as energy efficiency, Time to First Token latency and task performance, whilst assessing both AI agents and models for real-world network deployment.
GSMA Foundry is also supporting a AI Telco Troubleshooting Challenge, which calls on telecom operators, AI research and startups to submit advanced LLMs to perform root cause analysis and analyse network faults.
GSMA chief technology officer Alex Sinclair noted the collaboration will deliver AI capabilities specifically designed for telecommunications, addressing critical areas such as managing network faults and supporting operational decision-making for AI-driven network operations.
ZOOM OUT - The GSMA-Khalifa University collaboration builds on earlier groundbreaking work by Khalifa University’s 6G Research Centre, which developed the first version of TelecomGPT in mid-2024 through a joint project with Abu Dhabi’s Technology Innovation Institute. That initial model, trained on specialised telecom datasets, demonstrated superior performance over GPT-4, Llama-3 and Mistral in telecom-specific benchmarks including mathematical modelling tasks. The research established foundational evaluation frameworks for assessing LLM capabilities in telecommunications, addressing the sector’s lack of widely accepted benchmarks.
[Written and edited with the assistance of AI]
Read more about telecom GPT and related research:
GSMA releases telecom AI benchmarks ahead of MCW 2025 (Middle East AI News)
New leaderboard to support development of telecom LLMs (Middle East AI News)
Testing phase begins for TelecomGPT (Middle East AI News)
Abu Dhabi researchers create first-of-its-kind telecom LLM (Middle East AI News)