# CoreWeave Puts NVIDIA Vera Rubin NVL72 Into Production for Cloud Customers

By Simon Yoon

Canonical URL: https://www.tokenpost.com/news/technology/25883
Published: 2026-09-30T15:29:28.000Z
Updated: 2026-09-30T15:29:28.000Z
Section: Technology

> Cognition’s early tests found up to 4.8 times the total token throughput of NVIDIA’s GB200 NVL72 in software-engineering AI inference tasks.

CoreWeave has placed NVIDIA Vera Rubin NVL72 systems into production and opened access through its cloud platform, with Cognition’s tests showing up to 4.8 times the token throughput of NVIDIA’s GB200 NVL72.

The result came from real software-engineering AI inference tasks, where Cognition evaluated total token throughput. Cognition is the first customer to run production workloads on CoreWeave’s Vera Rubin NVL72 platform.

The NVL72 rack combines 72 Rubin GPUs with 36 Vera CPUs. Vera Rubin technology has also appeared in designs for larger, multi-rack AI infrastructure, including an [eight-rack system built around the NVL4 platform](<https://www.tokenpost.com/news/technology/23300>).

CoreWeave testing also showed that AI-agent sandboxes using NVIDIA Vera CPUs started more than three times faster. CoreWeave is making Vera Rubin NVL72 capacity available to customers through its cloud services.

## Links in this article

- [eight-rack system built around the NVL4 platform](https://www.tokenpost.com/news/technology/23300)