Telcos Turn to Open AI Models, NVIDIA Report Finds 89% Rate Them Important

Five reasons operators favor open models

NVIDIA argues that the appeal of open models goes beyond cost. In its view, they let telcos trust, control and customize AI across critical workloads, from autonomous networks to customer care. The company lists five strategic benefits.

• Access at lower cost: open models give access to frontier-level intelligence for less, so operators can keep closed models for the workloads where they add the most value. NVIDIA points to the Artificial Analysis Intelligence Index v4.3.2, which it says shows leading open models becoming more competitive on reasoning, coding, scientific and agentic tasks.
• Customization: open weights and training recipes can be fine-tuned with network, customer and industry data.
• Trust: greater visibility into and control over model artifacts and behavior lets models be evaluated, adapted and governed in line with regulations and business policies.
• Flexible deployment: teams can size and optimize open models for public clouds, private infrastructure and edge environments.
• Local services: telcos can host and fine-tune open models to offer locally adapted AI to enterprise and government customers.

Nemotron 3 Large Telco Model and a fine-tuning recipe

NVIDIA describes its Nemotron family as open models with frontier-level reasoning performance optimized for agentic workflows, plus speech capabilities for voice applications. The company says they come with open weights, training data and recipes.

NVIDIA also announced the Nemotron 3 Large Telco Model (LTM), which has 30 billion parameters. It was fine-tuned by AdaptKey on open source telecom datasets to improve accuracy on telecom-specific tasks. According to NVIDIA, it gives operators an open baseline that understands industry terminology and can reason through workflows such as network configuration and customer incident triage.

To help operators adapt the Nemotron 3 LTM and other open models with their own data, NVIDIA released the full recipe for an end-to-end fine-tuning pipeline. It covers adapting models to operator-specific networks, customers and procedures using NVIDIA NeMo open libraries.

SoftBank Corp. is one operator using open foundations this way. Rajeev Koodli, principal fellow of SoftBank Corp. and senior vice president of SB Telecom America, said the company uses them "extensively, including NVIDIA Nemotron models and others," to develop its SoftBank Large Telecom Model for uses such as network operations, design and overall management.

From models to production workflows

NVIDIA stresses that models alone are not enough to bring autonomous telecom operations safely into production. Operators need pipelines that prepare and protect data for fine-tuning, for example by anonymizing sensitive records and generating privacy-preserving synthetic datasets. They also need a platform that turns open models into governed, autonomous agentic workflows.

NVIDIA says it provides such an end-to-end platform, powered by NVIDIA AI Enterprise software and NVIDIA Agent Toolkit. It spans data pipelines, open models, agent orchestration, secure runtimes and simulation, and is backed by a partner ecosystem.

Andy Markus, chief data and AI officer of AT&T, said the future of AI "is not about choosing a single model; it's about intelligently matching every workload to the right combination of performance, cost and control." He added that with NVIDIA, AT&T can bring its model-choice strategy into production with the scale, reliability and governance its business requires.

Local language and national AI plans

NVIDIA also ties open models to national AI strategies. Operators in different countries can fine-tune them for local languages and industry needs, and host them on trusted platforms that customers use directly or build on with their own data and applications.

Indosat Ooredoo Hutchison, one of Indonesia's largest telecom operators, does this through its Sahabat-AI family of open source models. Chirag Sukhadia, its chief data and AI officer, said the approach lets the company "not only to adopt AI, but to build capabilities around it" and enable more Indonesians to create with AI on their own terms.

For telcos, the message from NVIDIA is that open models can serve as a base for both internal operations and new customer-facing services, provided they are paired with data preparation, governance and agent tooling. The 89% figure and the benchmark claims come from NVIDIA's own report and cited index, and the company has a commercial interest in the platform it describes.