European $100 billion fintech giant shifts to Zhipu GLM: AI coding bills cut by about 90%

Beating AI News Flash: Revolut, Europe's most valuable fintech unicorn, has made Zhipu GLM the default AI coding model for approximately 2,000 engineers. Tools like Claude Code and Codex have been relegated to backup options, mainly reserved for complex tasks.

David Abitbol, Head of AI Operations at Revolut, revealed that the company had previously set a $300 per person per month quota for Claude Code, but it contributed limited overall efficiency gains. In April of this year, the company began self-deploying GLM, offering engineers unlimited tokens.

Currently, Revolut uses approximately 56 NVIDIA B200 GPUs to run GLM 5.3, with a blended cost of about $0.30 per million tokens — only one-tenth of what it previously spent on Anthropic and OpenAI. According to internal company statistics, engineers' coding time metrics improved by approximately 60%, and the cycle time from task assignment to delivery improved by approximately 70%.

In the past, with limited tokens, engineers tended to save them for complex tasks. Now that they no longer need to ration usage, they have begun extensively using AI to search internal documents, process Jira tickets, and look up requirement specifications in Slack.

Only when tokens are cheap enough that there's no need to budget carefully can AI move from serving a handful of complex tasks to spanning the entire R&D workflow.