How Revolut Cut Its AI Coding Bill 90% by Self-Hosting an Open-Weight Model

Описание и применение

Революция в использовании ИИ: Revolut сократил расходы на 90 % и ускорил кодирование на 60 % за счёт самодостаточного развертывания модели GLM-5.3 на 56 GPU Nvidia B200.

Задача

Revolut’s AI coding assistants cost engineers up to $10,000/month with little measurable return.

Результат

Revolut reduced AI coding inference costs by ~90% (to $0.30/million tokens) and improved coding time by ~60%, lead time by ~70%, while maintaining 56 GPUs and achieving 76% throughput gain.

Рабочий процесс

• Deploy Gemini company-wide (early 2025)
• Introduce Claude Code with per-engineer budgets (early 2026)
• Begin self-hosting GLM-5.1 (April 2026)
• Upgrade to GLM-5.3 (~754B params) and tune inference
• Maintain 56 B200 GPUs while usage grows
• Use frontier models only for exceptions

Доказательства

Revolut did something almost no company of its size has tried: it ditched the frontier APIs as its default and moved its engineers onto a self-hosted, open-weight model.

Источник

The result, per Revolut's own numbers, was a roughly 90% cut in AI coding inference costs — plus coding that got about 60% faster.

Источник

GLM-5.3 (~754B total parameters, ~40B active per token), served from about 56 Nvidia B200 GPUs.

Источник