From NVIDIA: 2025-04-15 14:00:00

CoreWeave is now offering NVIDIA GB200 NVL72 systems to customers at scale, with Cohere, IBM, and Mistral AI already utilizing them for training and deploying AI models. Thousands of NVIDIA Blackwell GPUs are transforming raw data into intelligence with unprecedented speed. Leading AI companies are leveraging GB200 NVL72’s capabilities for cutting-edge model development.

Cohere is using Grace Blackwell Superchips to develop secure enterprise AI applications, experiencing up to 3x more performance in training for 100 billion-parameter models compared to previous GPUs. With further optimizations, Cohere is achieving higher throughput and shorter time to first and subsequent tokens for more efficient inference.

IBM is scaling to thousands of Blackwell GPUs on CoreWeave to train next-generation Granite models, providing state-of-the-art performance while maximizing safety, speed, and cost efficiency. CoreWeave’s deployment for IBM also includes the IBM Storage Scale System for high-performance storage for AI applications.

Mistral AI is using CoreWeave’s infrastructure to speed up the development of its language models, requiring large, high-performance GPU clusters with NVIDIA Quantum InfiniBand networking. Mistral AI saw a 2x improvement in performance for dense model training with NVIDIA GB200 NVL72, opening up new possibilities for model development and inference.



Read more at NVIDIA: Thousands of NVIDIA Grace Blackwell GPUs Now Live at CoreWeave, Propelling Development for AI Pioneers