What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
CoreWeave says NVIDIA Vera Rubin NVL72 is in limited availability on CoreWeave Cloud, with Cognition the first customer running production workloads. Cognition reported up to 4.8× total token throughput for its SWE-2 inference workload versus a GB200 NVL72 baseline. That is a workload-specific customer result, not a general performance guarantee.
What CoreWeave has put into production
In its September 30, 2026 announcement, CoreWeave said Vera Rubin NVL72 was in limited availability on CoreWeave Cloud. The company reported hundreds of Rubin GPUs deployed across multiple regions and said customers were beginning to onboard workloads. Cognition was identified as the first customer running production agentic AI workloads on the platform. CoreWeave’s announcement describes a Vera Rubin NVL72 rack as pairing 72 Rubin GPUs with 36 Vera CPUs.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
The NVIDIA Vera Rubin AI Chip: The CES 2026 Launch, Rubin GPU Innovations, and AI Breakthroughs for... | $20.99 | Buy on Amazon |
“Limited availability” is the key qualification: the announcement does not say that capacity is generally available to every customer. CoreWeave says customers can use the same operating model and tooling as they do with existing GB200 and GB300 NVL72 fleets, but the announcement does not publish a price comparison.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsWhat Cognition’s 4.8× result measures
Cognition’s Silas Alberti, SVP Research & Founding Team, described agentic coding as a workload requiring long contexts, high concurrency and rapid reasoning. CoreWeave’s investor release attributes an independent benchmark run on CoreWeave Cloud to Cognition: up to 4.8× total token throughput for SWE-2 inference compared with GB200 NVL72. The company’s blog chart characterizes the inference comparison at matched interactivity and per GPU. The investor announcement does not provide enough methodological detail to extend the ratio to other models, configurations or latency targets.
#1 Best Overall
The number is specifically an upper-bound result for Cognition’s SWE-2 inference workload. It should not be read as 4.8× faster completion time, a general GPU speedup, or a benchmark replicated across unrelated workloads.
Reinforcement-learning result is a separate metric
CoreWeave also reports Cognition measured 3.8× output-token throughput for reinforcement-learning workloads versus GB200 NVL72. That is a different workload and metric from the 4.8× total-token-throughput inference result; the two figures should not be combined into a single performance claim.
What NVIDIA Vera CPU adds
NVIDIA says CoreWeave will also offer Vera CPU, a processor platform aimed at CPU-intensive work associated with agentic AI. NVIDIA’s September 30 account describes a CoreWeave rack with 128 Vera CPUs and 11,264 cores. It reports more than 3× faster agent-sandbox startup in CoreWeave testing and a 1.7× performance gain across passing Terminal-Bench tasks. These are results reported in NVIDIA’s account, not independently audited findings in the material reviewed.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Do not confuse that CoreWeave setup with the separate Vera CPU rack configuration NVIDIA describes elsewhere: that rack integrates 256 liquid-cooled Vera CPUs and is intended to support more than 22,500 concurrent CPU environments. The 256-CPU figure is not a specification for the 128-CPU CoreWeave rack. NVIDIA founder and CEO Jensen Huang framed the platform’s role this way: “The CPU is no longer simply supporting the model; it’s driving it.” NVIDIA’s Vera CPU announcement provides that distinct rack description.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What CoreWeave Forge is
NVIDIA describes CoreWeave Forge as a connected environment for training, evaluating and improving models and agents. The announced components include Weights & Biases, OpenPipe post-training expertise and the open-source marimo notebook project. The proposition is an integrated workflow for iterative model and agent development; the announcement does not establish public pricing or a general availability schedule for every Forge component. NVIDIA’s CoreWeave post also says CoreWeave’s multi-rack systems can connect hundreds of accelerators using Spectrum-X Ethernet, as covered in CoreWeave’s multi-rack announcement.
Quick Recap
What infrastructure buyers should take from the announcement
- Availability: Rubin NVL72 is described as in limited availability on CoreWeave Cloud, with customer onboarding underway.
- Workload fit: Cognition’s reported throughput figures concern its SWE-2 inference and reinforcement-learning workloads; they do not establish results for another team’s software or serving targets.
- Scale and operations: CoreWeave describes a 72-GPU, 36-CPU rack and says customers can operate it with tooling used for its GB200 and GB300 NVL72 fleets. Its multi-rack announcement discusses Spectrum-X Ethernet connections across hundreds of accelerators.
- CPU-intensive agent work: NVIDIA’s reported sandbox-startup and Terminal-Bench results indicate the intended Vera CPU use case, but they remain company-reported tests.
- Cost: The cited announcements do not state a public price or a cost comparison with GB200 NVL72.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
