CoreWeave Leads Cloud Providers in MLPerf Inference V6.1 Performance with NVIDIA Blackwell Ultra
CoreWeave, Wednesday, September 16th, 2026
CoreWeave posts leading cloud-provider results in MLPerf Inference v6.1 using NVIDIA Blackwell Ultra systems.
CoreWeave reported MLPerf Inference v6.1 results placing it ahead of other cloud providers on NVIDIA Blackwell Ultra hardware.
The post covers the benchmarked workloads and the infrastructure tuning behind the numbers, including networking, storage, and scheduler configuration.
It argues that raw accelerator specifications do not determine delivered inference performance and that platform engineering accounts for a substantial share of the result. CoreWeave presents the submission as an independently verifiable basis for comparing AI cloud providers.