NVIDIA Vera Rubin Debuts in MLPerf, Leads Training and Endpoints
Start here
NVIDIA's Vera Rubin platform made its MLPerf debut, and NVIDIA says it was the only submitter with results on every MLPerf Training v6.1 benchmark. According to NVIDIA, Vera Rubin NVL72 delivered up to 2x more performance per GPU than GB300 NVL72, and on the DeepSeek-V3 benchmark it scaled from 256 to 1,024 GPUs at about 97% efficiency, cutting training time from 16 minutes to about four. In its first MLPerf Endpoints submission, NVIDIA reports Vera Rubin NVL72 showed more than 10x the throughput per watt of GB300 NVL72 on a provisional Kimi K3 agentic result at the same response speed, measured at a shared minimum end-to-end average interactivity of 50 tokens per second per user. Those Vera Rubin Endpoints results are provisional and pending MLCommons verification and peer review; NVIDIA says the GB300 results were verified by MLCommons.
Generated from reports · updated 34 min ago
Report timeline
Follow the reports to see the event from different sides.
- NVIDIA BlogPicksNVIDIA Vera Rubin debuts in MLPerf, sweeping Training v6.1 and leading Endpoints agentic inference
NVIDIA's Vera Rubin platform made its MLPerf debut with results across training, post-training and agentic inference, and NVIDIA was the only submitter with results on every MLPerf Training v6.1 benchmark. Vera Rubin NVL72 delivered up to 2x more performance per GPU than GB300 NVL72, and on the DeepSeek-V3 benchmark it scaled from 256 to 1,024 GPUs at about 97% efficiency, cutting training time from 16 minutes to about four. In its first MLPerf Endpoints submission, Vera Rubin NVL72 showed more than 10x the throughput per watt of GB300 NVL72 on a provisional Kimi K3 agentic result at the same response speed, though those results are pending MLCommons verification.
Heat for this event
Not enough continuous observations yet to draw a trend.