Case Study: Stability AI achieves faster, more affordable AI training with WEKA

Stability AI - Customer Case Study

Stability AI, an open-source generative AI company, needed faster and more cost-effective AI training in the cloud. As model demand grew, its legacy Lustre-based storage environment created performance bottlenecks, low GPU utilization, and unpredictable storage costs that threatened speed to market.

Stability AI worked with Weka to deploy WEKA Data Platform in Converged Mode on AWS, running storage and training on the same compute, network, and storage resources. The result was 93% GPU utilization, a 95% reduction in storage cost per TB, and a 35% improvement in training time, cutting about three weeks from a 60-day cycle.

Stability AI

Richard Vencu

MLOps Lead