NVIDIA Makes the Case for Open Models in Telecom and Announces a 30B Telco Model

In an October 6 post, NVIDIA says telecom operators are choosing open models for more than cost: they offer operators greater ability to trust, control and customize AI for workloads including autonomous networks and customer care. NVIDIA cites its State of AI in Telecommunications report, in which 89% of respondents consider open-source models and software important to their company’s AI strategy.

The company says open weights and training recipes allow operators to fine-tune models using network, customer and industry data, while deployment can span public clouds, private infrastructure and edge environments. NVIDIA describes its NVIDIA Nemotron family as providing reasoning for agentic workflows and speech capabilities for voice applications, with open weights, training data and recipes.

SoftBank Corp.’s Rajeev Koodli says the company uses open foundations, including NVIDIA Nemotron models, to develop its SoftBank Large Telecom Model for telecom-specific uses such as network operations, design and overall management. NVIDIA also announced the 30-billion-parameter Nemotron 3 Large Telco Model, which AdaptKey fine-tuned on open-source telecom datasets to improve accuracy on telecom-specific tasks.

NVIDIA says it released an end-to-end fine-tuning recipe built with NVIDIA NeMo open libraries so operators can adapt models to their own networks, customers and procedures. The announcement builds on NVIDIA’s earlier telecom-AI coverage, which described synthetic data as one way to protect sensitive information while preparing models for telecom operations.