NVIDIA Vera Rubin NVL72 Delivers Leading Performance in MLPerf Inference V6.1 Debut
NVIDIA, Wednesday, September 16th, 2026
Vera Rubin NVL72 posts leading results in its MLPerf Inference v6.1 debut submission.
NVIDIA reports that the Vera Rubin NVL72 rack-scale system delivered leading performance in its first MLPerf Inference v6.1 submission.
The post breaks down results across benchmark workloads and explains how NVLink-connected rack-scale design improves throughput on large models.
NVIDIA attributes gains to the combination of new silicon, higher memory bandwidth, and software stack optimization.
Per-watt and per-rack figures are highlighted to support AI factory efficiency claims.