CoreWeave makes Nvidia Vera Rubin NVL72 available, with Cognition as first production customer

CoreWeave announced Vera Rubin NVL72 availability on its cloud at its Fully Connected conference, naming Cognition as the first customer running production workloads on the system. Cognition reports up to 4.8 times the total token throughput of a GB200 NVL72 baseline for its SWE-2 inference work. CoreWeave also cited its own measurements showing 10 times more tokens per megawatt than GB200 on DeepSeek R1.

• Cognition is the first customer running production workloads on Vera Rubin NVL72
• Cognition reports up to 4.8x total token throughput versus GB200 NVL72 (SWE-2 inference)
• CoreWeave's own benchmark: 10x tokens per megawatt versus GB200 NVL72 on DeepSeek R1