Cacheon (SN14) boosts Affine (SN120)’s King model inference throughput by 40.9% in two weeks, reaching 528.9 tokens per second as Affine becomes its first commercial customer. […]