Video Captioning at Scale: 600 TB in 95 Minutes With Anyscale on CoreWeave
CoreWeave, Friday, August 28th, 2026
CoreWeave and Anyscale captioned 600 TB of video in 95 minutes on CoreWeave infrastructure.
CoreWeave describes a large-scale video captioning workload run with Anyscale on CoreWeave infrastructure, processing 600 terabytes of video in 95 minutes. Video captioning at that volume is a demanding pipeline problem as much as a model problem, since data movement, decoding and scheduling determine throughput once the model itself is fast enough.
The post covers how the workload was structured across the cluster and what made the throughput achievable, aimed at teams building large multimodal data preparation pipelines.