Revolut cuts AI coding costs 90% via self-hosted models
Revolut reduced AI inference costs by 90% by switching to self-hosted models.
Revolut reduced its AI coding inference costs by roughly 90 percent in less than a year. The company achieved these savings by making self-hosted models its default instead of relying on commercial services like Claude. David Abitbol, ops and infrastructure lead at Revolut, spoke about the cost improvements at the AI Engineer conference in New York last month.
Revolut rolled out AI coding tools to roughly 2,000 engineers. The company saw a 60 percent improvement in coding time and a 70 percent improvement in lead time after the rollout. Engineers previously spent up to $10,000 a month on AI tools with limited ROI. Revolut launched its AI tools with strict budget limits due to cost concerns.
The company initially adopted the GLM-5.1 model before moving to GLM-5.3. Revolut now serves its AI models using 56 Nvidia B200 GPUs. Self-hosting GLM 5.1 cost about $0.30 per million input/output tokens. This was about a tenth of using OpenAI's Codex or Anthropic's Claude Code. The biggest gains from AI models at Revolut came from handling debugging and Jira and Confluence tickets.
Sources
Every sentence of this story was checked against facts quoted from these articles. How we make the news
Spotted an error? Tell us
