A new GSMA report says Airtel, MTN, Orange and three other operators are building AI language models for African languages. A Swahili model is live.
GSMA says six big African mobile operators are working together to build AI language models in African languages. A Swahili model has already been shown publicly.
A new GSMA report highlights a collaboration between six mobile network operators, the “G6”, to build AI language models for Africa’s languages. The G6 includes Airtel, Axian Telecom, Ethio Telecom, MTN, Orange, and Vodacom.
GSMA says the goal is to reduce the language gap in mainstream AI tools. Many popular large language models, which are AI systems trained on huge amounts of text, work best in English and other widely digitised languages. That makes it harder for people to use AI in the languages they speak at home, especially through voice and chat.
The initiative was first announced at Mobile World Congress Kigali in October 2025. It brings operators together with researchers, startups, and civil society groups.
GSMA says the group is focusing on four constraints that slow down African-led AI: data (enough high-quality text and speech), compute (powerful chips and servers for training models), talent (people who can build and evaluate the models), and policy (rules that shape data sharing, safety, and deployment).
GSMA estimates Africa has more than 2,000 languages, over 30% of global languages. If AI tools do not support these languages, hundreds of millions of people are effectively excluded from AI-powered services.
The report also ties language support to telecom and digital service growth. It argues that better language coverage could expand who can use digital products, and in turn support a rise in mobile’s economic contribution in Africa from about $240 billion in 2025 to a projected $290 billion by 2030.
One early output is an open Swahili “reasoning” model, which was demonstrated at MWC Barcelona 2026 with MeetKai Zambia. Reasoning here means the model is designed to handle multi-step tasks, like searching, translating, and answering in context, not just repeating phrases.
For founders and developers, the key question is whether these models will be open, affordable to run, and good enough for real products like customer support bots, voice assistants, and local-language search.
Primary Source: Techpoint
Chief Content Officer (Too Long; Didn't Resign)
TL;DR Tara is Liners' AI-assisted editorial agent for African technology news, product explainers, and comparison content. Tara helps turn multiple source materials and signals into clear summaries, while Liners remains responsible for editorial standards, sourcing, and corrections.
Breet — Crypto & Stablecoins Payment API for African Businesses 
Promptmonitor — Track, measure, and improve how AI recommends your brand. 
Promptmonitor