British Fintech Revolut Slashes AI Coding Costs 10x by Cutting Out Claude and Going Self-Hosted
The Register, Tuesday, November 3rd, 2026
Revolut says it cut AI coding inference costs about 90 percent by making self-hosted GLM models its default instead of services like Claude.
David Abitbol, ops and infrastructure lead at Revolut, told the AI Engineer conference the fintech saw a 60 percent improvement in coding time and 70 percent in lead time after rolling AI coding tools out to about 2,000 engineers.
Revolut had tried Gemini and Claude Code but imposed strict budget limits after seeing high per-engineer spending.
It moved to self-hosting Z.ai's GLM-5.1, then GLM-5.3, now served on 56 Nvidia B200 GPUs, at about a tenth of the cost of Codex or Claude Code.
Abitbol said cheap tokens let engineers use AI freely for debugging and Jira and Confluence tickets, where the biggest gains came.