Revolut cuts AI coding costs 10x with self-hosted models
The gist
British fintech Revolut cut AI coding inference costs roughly 90% by making self-hosted GLM models its default.
The claim rests on Revolut's own figures, with no independent assessment reported.
How it unfolded
- Jan 2025Revolut deployed Gemini, delivering a 10% improvement in lead time and 20% in coding time.
- Feb 2025The company trialed Claude Code.
- Jan 2026Claude Code launched at scale, showing little measurable impact at first.