NVIDIAEventDevelopersSeptember 16, 2026

NVIDIA Vera Rubin NVL72 debuts in MLPerf Inference v6.1

Vera Rubin NVL72's first MLPerf Inference preview submission delivers up to 3.7x higher throughput than GB300 NVL72 on Qwen3-VL, using vLLM. A 288-GPU GB300 NVL72 submission across four racks hit 99% scaling efficiency.

1 source

More stories today

Open the live feed