Revolut says self-hosted AI coding models cut its inference costs by about 90 percent

Revolut reports that moving its AI coding workload to self-hosted GLM-5.x models, which it serves on 56 Nvidia◼ B200 GPUs, cut inference costs by roughly 90 percent compared with commercial services such as Claude◆. The firm says coding time improved by 60 percent and lead time by 70 percent across about 2,000 engineers. It tuned quantisation, disaggregated prefill and decode, and context length, and uses 256,000 tokens as its practical context window.

For IT people · Written by AI from 1 source · Wrong?