• AT&T has unveiled OTel 2.0
• It’s the latest open telco AI model being trained by AT&T and AMD with help from the GSMA
• The model is the ‘best and largest’ model trained specifically for use by telcos, reckons the GSMA

Here’s a development of vital importance to network operators with AI-native telco aspirations. 

Earlier this year, the GSMA launched Open Telco AI, an initiative that aims to address the fact that generic AI frontier models aren’t tuned for the specific needs of telecom network operators by fostering the development of telco-grade AI models. 

Louis Powell, director of AI initiatives at the GSMA, explained to TelecomTV during MWC26 that AT&T and AMD were laying the foundations for the initiative: That collaboration resulted in the development by AT&T of a family of open telco models called OTel that were trained on a specialised telco-specific dataset curated by GSMA and its collaborators, including telecom operators, network equipment providers and academia. 

Now AT&T has unveiled OTel 2.0, which has been trained, with help from the likes of Red Hat, Dell Technologies and Microsoft Azure, using more than 400 billion tokens on AMD GPUs. “We’re training these models to provide the telco industry with more flexibility and better optimisation,” noted Andy Markus, AT&T’s chief data and AI officer, in this blog. 

OTel models “drive down token costs by providing model options that are suited to our tasks, and can run on right-sized hardware, either on premises or in the cloud,” he noted. 

In a separate blog, the GSMA’s Powell describes OTel 2.0 as “the best and largest post-trained open model built specifically for the telecoms industry.” 

He added: “Domain-adapted models are highly accurate and can be significantly smaller in size. Just as healthcare, financial services and manufacturing are developing domain-specific AI approaches, the telecoms industry needs models trained on its unique standards, protocols and operating environments. This is not just about accuracy it is about enterprise requirements; reducing costs and maintaining control by deploying models across clouds and on-premise as needed. Accuracy and requirements together are vital for operators deploying telco-specific use cases, like network troubleshooting, product development, network configuration and more.” 

There’s a lot more insight in both blogs, which are recommended reading, and it’s worth remembering that the GSMA is one of the founding partners of TelecomTV’s AI-native Telco Accelerator (ANTA) initiative, so these developments will also bring critical insights to the ANTA community.  

One particular development of note from the Markus blog is how AT&T has adapted its internal AI workflow processes to connect each workload with the optimum AI model which, as Markus notes, doesn’t need to be “the latest and greatest models”, as many tasks “can be handled by lower-cost models without sacrificing performance.”  

To align tasks with the most suitable model, AT&T developed an AI gateway “that goes beyond simple model selection: It uses cache-aware routing to intelligently match each task to the most cost-effective model – without compromising on quality. It can even change models partway during multi-turn sessions. At each turn, the gateway weighs speed and cost with the expected quality of the output, then routes the prompt to the best model. The result: We’re reducing AI costs as much as 90%, already saving millions.” 

That’s the kind of impact that will interest every AI-native telco.   

- Ray Le Maistre, Editorial Director, TelecomTV