An open guide by NVIDIA and Tech Mahindra shows operators how to fine-tune domain-specific reasoning models and build agents that execute NOC workflows using the NeMo-Skills pipeline for telco learning.

NVIDIA released an open Nemotron-based large telco model (LTM) built on Nemotron 3 foundations and fine-tuned with telecom data by AdaptKey AI, designed to identify faults, plan remediation with rollback, and validate changes before deployment, and it is open to operators for on-premises deployment with full training transparency.

NVIDIA’s blueprint for network configuration is being adopted by Cassava Technologies and NTT DATA to enable agent-driven monitoring, controlled change application, and impact assessment with safe rollback.

Telenor Group will be the first to deploy the updated blueprint with BubbleRAN to improve 5G network performance for Telenor Maritime, testing agentic orchestration and closed-loop operations in challenging connectivity scenarios.

Multi-agent orchestration is being enhanced with NVIDIA NeMo Agent Toolkit (NAT) and BubbleRAN’s BAT tools, integrated into BubbleRAN’s Opti-Sphere platform to coordinate monitoring, configuration, and validation across containers and workloads.

Operators are adopting a production-ready telco blueprint for multi-vendor RAN with three specialized agents—monitoring/recommendation, execution, and validation/rollback—to govern changes with governance and safe rollback.

A closed-loop autonomy blueprint for intent-driven RAN energy efficiency in 5G uses VIAVI’s AI RSG to generate synthetic data and an energy-planning agent to validate policies before live deployment.

Further details and demonstrations are anticipated at Mobile World Congress in Barcelona in early March 2026.

The GSMA Open Telco AI initiative will release the LTM, implementation guide, and agentic AI blueprints as open resources.

NTT DATA is applying the blueprint to improve traffic regulation after outages, with an AI agent making admission-control decisions to manage surge traffic in a Tier-1 Japanese network.

NVIDIA is advancing autonomous networks in telecom by developing agentic AI and telco-specific reasoning models to move from automation to autonomous operations, highlighting operator needs for intent understanding, option weighing, and safe action.

The LTM is open-source to operators, ensuring transparency of training data and methods, enabling deployment within security boundaries and customization with internal data.