NVIDIA Open Source 6G Tools Top 1 Million Downloads, Accelerate AI Network Research
Open source NVIDIA Aerial and Sionna adoption rose 10x, topping 1 million downloads as researchers cut robot latency, power use and 6G simulation times.
Summary
On November 6, 2026, one year after NVIDIA open sourced Aerial, global adoption of NVIDIA AI Aerial and NVIDIA Sionna was up 10x. NVIDIA’s open source 5G/6G tools surpassed 1 million downloads and were used by more than 8,000 research and academic institutions. The software links simulation and AI training to over-the-air experiments, supporting AI-native 6G that combines connectivity, computing and sensing for billions of people, machines and agents.
King’s College London paired a camera-equipped robot arm with NVIDIA Jetson AGX Orin and Sionna for software-defined 5G and edge AI. Goal-oriented semantic communications sent compact object and spatial descriptions instead of full images. In fault-injected fruit-packing and cube-stacking tests, task time fell up to 52.6% versus periodic image transmission, success improved, failed actions were recovered, and digital twin configurations checked collisions. SoftBank integrated a neural receiver into NVIDIA Aerial CUDA-Accelerated RAN, or ACAR. Using a commercial 5G smartphone under controlled interference, it beat a conventional receiver in all 15 settings, with roughly 60% maximum uplink gain. Using NVIDIA Aerial Testbed, ETH Zurich cut block error rates by more than one-third versus selected pretrained receivers through local tuning.
University of Leeds researchers used 10-minute demand forecasts to sleep base stations only when neighbors could maintain service. A portable NVIDIA DGX Spark-powered 5G network-in-a-box and digital twin cut RAN power 45.6% versus full power while preserving connectivity. The Centre Tecnològic de Telecomunicacions de Catalunya, or CTTC, added GPU-accelerated ACAR radio models to open source 5G-LENA, cutting early simulation runtime from about three hours to 10 to 20 minutes while preserving expected statistics and performance. AI Aerial and Sionna are on GitHub, with further work planned for 6G Developer Sessions at NVIDIA GTC Washington, D.C.
Positives
- NVIDIA’s open source 5G/6G tools exceeded 1 million downloads as AI Aerial and Sionna adoption grew 10x in one year.
- King’s College London reduced robot task completion time by up to 52.6% while improving success and recovery from failed actions.
- SoftBank’s neural receiver increased uplink throughput at all 15 tested interference settings, with a maximum gain of roughly 60%.
- ETH Zurich reduced block error rates by more than one-third by tuning neural receivers with local measurements.
- University of Leeds researchers cut RAN power consumption 45.6% while maintaining device connectivity.
- CTTC reduced network simulation runtime from about three hours to 10 to 20 minutes while preserving expected behavior.
Risks & concerns
- Interference and changing radio conditions can cause decoding errors and retransmissions that delay urgent field imagery.
- Growing 6G device counts and AI workloads will make network energy management increasingly important.
- SoftBank’s neural receiver evidence remains limited to controlled interference tests across 15 settings.
- CTTC’s three-hour to 10-to-20-minute simulation speedup comes from an early evaluation.
- King’s College London measured robot gains in fruit-packing and cube-stacking experiments with deliberately introduced faults.