CoreWeave, NVIDIA and IBM Submit Largest-Ever MLPerf Results on NVIDIA GB200 Grace Blackwell Superchips

Rhea-AI Summary

CoreWeave (CRWV), NVIDIA, and IBM achieved record-breaking MLPerf Training v5.0 results using 2,496 NVIDIA GB200 Blackwell GPUs on CoreWeave's AI cloud platform. The submission represents the largest NVIDIA GB200 NVL72 cluster ever benchmarked, 34x larger than other cloud provider submissions. The cluster completed training the Llama 3.1 405B model in just 27.3 minutes, demonstrating 2x faster training performance compared to similar-sized clusters. CoreWeave's achievement showcases their platform's capability to deliver superior AI workload performance, offering customers significant advantages in model development speed and cost efficiency. The company's leadership in AI infrastructure is further validated by its Platinum tier ranking in SemiAnalysis' ClusterMAX and strong performance in both MLPerf Inference v5.0 and Training v5.0 benchmarks.

Positive

• Record-breaking MLPerf submission with largest-ever NVIDIA GB200 NVL72 cluster (2,496 GPUs)
• 2x faster training performance compared to similar cluster sizes
• 34x larger cluster than other cloud provider submissions
• Completed Llama 3.1 405B model training in just 27.3 minutes
• Platinum tier ranking in SemiAnalysis' ClusterMAX

Negative

• None.

Submission with nearly 2,500 NVIDIA GB200 GPUs achieved breakthrough results on most complex benchmarking model

LIVINGSTON, N.J., June 4, 2025 /PRNewswire/ -- CoreWeave (Nasdaq: CRWV), in collaboration with NVIDIA and IBM, delivered the largest-ever MLPerf® Training v5.0 submission on NVIDIA Blackwell, using 2,496 NVIDIA Blackwell GPUs running on CoreWeave's AI-optimized cloud platform. This submission is the largest NVIDIA GB200 NVL72 cluster ever benchmarked under MLPerf, 34x larger than the only other submission from a cloud provider highlighting the large scale and readiness of CoreWeave's cloud platform for today's demanding AI workloads.

The submission achieved a breakthrough result on the largest and most complex foundational model in the benchmarking suite–Llama 3.1 405B–completing the run in just 27.3 minutes. When compared against submissions from other participants across similar cluster sizes, CoreWeave's GB200 cluster achieved more than 2x faster training performance. This result highlights the significant performance leap enabled by the GB200 NVL72 architecture and the strength of CoreWeave's infrastructure in delivering consistent, best-in-class AI workload performance.

"AI labs and enterprises choose CoreWeave because we deliver a purpose-built cloud platform with the scale, performance, and reliability that their workloads demand," said Peter Salanki, Chief Technology Officer and Co-founder at CoreWeave. "These MLPerf results reinforce our leadership in supporting today's most demanding AI workloads."

These results matter because they translate directly to faster model development cycles and an optimized Total Cost of Ownership. For CoreWeave customers, that means cutting training time in half, scaling workloads efficiently, and training or deploying their models more cost-effectively by leveraging the latest cloud technologies, months before their competitors. With leading submissions for both MLPerf Inference v5.0 and Training v5.0 benchmarks and the sole cloud provider ranked in the Platinum tier of SemiAnalysis' ClusterMAX, CoreWeave sets the standard for AI infrastructure performance across the entire cloud stack.

About CoreWeave

CoreWeave, the AI Hyperscaler™, delivers a cloud platform of cutting-edge software powering the next wave of AI. The company's technology provides enterprises and leading AI labs with cloud solutions for accelerated computing. Since 2017, CoreWeave has operated a growing footprint of data centers across the US and Europe. CoreWeave was ranked as one of the TIME100 most influential companies and featured on Forbes Cloud 100 ranking in 2024. Learn more at www.coreweave.com.

The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information.

 

View original content:https://www.prnewswire.com/news-releases/coreweave-nvidia-and-ibm-submit-largest-ever-mlperf-results-on-nvidia-gb200-grace-blackwell-superchips-302473361.html

SOURCE CoreWeave

FAQ

What did CoreWeave achieve in the MLPerf Training v5.0 benchmarks?

CoreWeave achieved the largest-ever MLPerf Training v5.0 submission using 2,496 NVIDIA GB200 Blackwell GPUs, completing Llama 3.1 405B model training in 27.3 minutes with 2x faster performance than similar clusters.

How does CoreWeave's NVIDIA GB200 cluster compare to other cloud providers?

CoreWeave's cluster is 34x larger than the only other cloud provider submission, demonstrating superior scale and AI workload capabilities.

What are the benefits for CoreWeave (CRWV) customers from these MLPerf results?

Customers can cut training time in half, scale workloads efficiently, and train or deploy models more cost-effectively using the latest cloud technologies months ahead of competitors.

What certifications or rankings does CoreWeave hold in AI infrastructure?

CoreWeave is ranked in the Platinum tier of SemiAnalysis' ClusterMAX and has leading submissions in both MLPerf Inference v5.0 and Training v5.0 benchmarks.

How long did it take CoreWeave to train the Llama 3.1 405B model?

CoreWeave completed the training of Llama 3.1 405B model in just 27.3 minutes using their NVIDIA GB200 GPU cluster.