CoreWeave announced that its cloud platform now offers NVIDIA Vera Rubin NVL72 systems, featuring Spectrum-X 102.4T Ethernet networking. AI company Cognition (developer of Devin AI) is among the first clients, with early tests showing Vera Rubin NVL72 delivering up to 4.8 times higher total token throughput than GB200 NVL72 in SWE-2 inference workloads. Additionally, CoreWeave will provide the NVIDIA Vera CPU, designed for AI agents, which achieved over 3 times faster agent sandbox launch speeds in testing. CoreWeave also introduced CoreWeave Forge, an integrated environment for training, evaluating, and refining models and agents on NVIDIA accelerated computing platforms.