CoreWeave Sets New AI Benchmark with NVIDIA GB200 Superchips

CoreWeave has achieved a new record in AI inferencing benchmarks using NVIDIA GB200 Grace Blackwell Superchips, announced in a press release. The company reported delivering 800 tokens per second (TPS) on the Llama 3.1 405B model, one of the largest open-source models, using a CoreWeave instance equipped with two NVIDIA Grace CPUs and four NVIDIA Blackwell GPUs.

Additionally, CoreWeave submitted results for NVIDIA H200 GPU instances, achieving 33,000 TPS on the Llama 2 70B model, marking a 40% improvement over previous NVIDIA H100 instances. These achievements underscore CoreWeave's position as a leading provider of cloud infrastructure services optimized for AI applications.