CoreWeave says Vera Rubin NVL72 made 10 times the tokens per megawatt

CoreWeave said on Nov. 19 that Vera Rubin NVL72 generated 10 times as many tokens per second per megawatt as GB200 NVL72 on DeepSeek R1 at a matched interactivity target. Its product page lists 800,000 versus 80,000 tokens per second per megawatt at 150 tokens per second per user, and the post calls the result a starting point on a rack of 72 Rubin GPUs and 36 Vera CPUs linked by 260 TB/s of NVLink 6. The figure is CoreWeave's measurement, not a new Nvidia specification.

/ Sources

/ About this story

Compiled by Venture Atlas from the sources above, using automated AI-assisted research. This is a summary of reporting published elsewhere, not original reporting - follow the source links for the full account. See our editorial standards.

Something wrong here? Email apostructura@gmail.com and we will fix it.

/ Related