NVIDIA reported that 89 percent of respondents in its latest State of AI in Telecommunications survey consider open source models and software important to their company's artificial intelligence strategy. The company outlined a fivefold strategic incentive for open architectures, citing lower deployment costs, deeper operational customization, model governance, multi-environment deployments across edge and cloud infrastructure, and the ability to offer localized enterprise services.
Partner AdaptKey fine-tuned NVIDIA's newly announced 30-billion-parameter Nemotron 3 Large Telco Model on open-source telecommunications datasets. The model serves as an open baseline designed to process industry terminology and assist with network configuration and customer incident triage. To support operator adjustments, NVIDIA published the complete pipeline recipe for adapting open models to network records and operational procedures using its NeMo open libraries.
SoftBank Corp. is deploying open foundations, including the NVIDIA Nemotron family, to build its SoftBank Large Telecom Model. Rajeev Koodli, principal fellow of SoftBank Corp. and senior vice president of SB Telecom America, said open models allow the company to apply accumulated network knowledge to foundation models for network design, operations, and overall management.
Andy Markus, chief data and AI officer of AT&T, said the operator's strategy centers on matching workloads to performance, cost, and control rather than committing to a single model. To take open models into live environments, NVIDIA is providing an operational stack based on NVIDIA AI Enterprise software and the NVIDIA Agent Toolkit, which covers synthetic data generation, sensitive record anonymization, agent orchestration, and secure runtimes.
Indosat Ooredoo Hutchison is using open source models to build Sahabat-AI, an initiative tailored to Indonesian languages and regional cultural requirements. Chirag Sukhadia, chief data and AI officer of the Indonesian operator, said open architectures allow local providers to build sovereign technical capabilities. Independent evaluations, including the Artificial Analysis Intelligence Index v4.3.2, show leading open models are closing the performance gap with proprietary models in reasoning, coding, and agentic tasks.
