Report: GKE Inference Gateway delivers up to 92% faster AI responses

Google Kubernetes Engine Inference Gateway routes LLM requests using real-time model metrics and prefix-cache-aware, model-aware routing to reduce accelerator recomputation and latency. An independent benchmark reports 15.7% higher throughput, 92.8% shorter time to first token, and 62.6% lower inter-token latency versus round-robin load balancing.