Serverless GPU Inference Scaling Breakthrough

Modal achieved serverless GPU inference scaling improvements, reducing replica spin-up time from kiloseconds to tens of seconds, enabling practical variable-load AI inference workloads.