CoreWeave, the AI Hyperscaler™, today announced its MLPerf v5.0 results, setting a new industry benchmark in AI inference with NVIDIA GB200 Grace Blackwell Superchips. Using a CoreWeave instance with NVIDIA GB200, featuring two NVIDIA Grace CPUs and four NVIDIA Blackwell GPUs, CoreWeave delivered 800 tokens per second (TPS) on the Llama 3.1 405B model¹—one of the largest open-source models.

"CoreWeave is committed to delivering cutting-edge infrastructure optimized for large-model inference through our purpose-built cloud platform," said Peter Salanki, Chief Technology Officer at CoreWeave. "These benchmark MLPerf results reinforce CoreWeave's position as a preferred cloud provider for leading AI labs and enterprises."

CoreWeave also submitted new results for NVIDIA H200 GPU instances. It achieved 33,000 TPS on the Llama 2 70B model, representing a 40 percent improvement in throughput over NVIDIA H100 instances