CoreWeave makes Nvidia Vera Rubin NVL72 available, Cognition is early customer

CoreWeave is now offering the Nvidia Vera Rubin NVL72 system on its cloud platform, with Cognition as the first customer using it since early September. Cognition reported up to a 4.8x increase in total token throughput for its SWE-2 inference workloads. CoreWeave also plans to offer the Vera CPU standalone and launched CoreWeave Forge.

• Cognition has used Vera Rubin NVL72 since early September 2026
• Up to 4.8x increase in token throughput on SWE-2 workloads
• NVL72 packs 36 CPUs and 72 GPUs with full liquid cooling
• Vera CPU offered standalone: 128 CPUs and 11,264 cores per rack