Google Cloud C4 Brings a 70% TCO improvement on GPT OSS with Intel and Hugging Face
Intel and Hugging Face demonstrate real-world gains running GPT OSS on Google Cloud C4 VMs with Granite Rapids, achieving up to 1.7x lower TCO and 1.4–1.7x higher TPOT throughput per vCPU/dollar versus C3. The work optimizes MoE expert execution to avoid redundant computation, improving utilization on large models.