tech
Revolut Cuts AI Coding Inference Costs by About 90% Using Self-Hosted Models
Revolut has cut AI coding inference costs by about 90% by switching to self-hosted models instead of commercial services.
Key points
- Revolut says self-hosted models are now its default, instead of relying primarily on commercial services like Claude.
- Self-hosting GLM 5.1 at $0.30 per million input/output tokens cost about a tenth of OpenAI's Codex or Anthropic's Claude Code.
- Revolut serves GLM-5.3 on 56 Nvidia B200 GPUs, and the model is text only, so it isn't ideal for UI work.
Why it matters. Revolut's experience suggests that self-hosting open models may sharply lower AI coding costs, though the company has not replaced GPT or Claude entirely.
Revolut says it has cut its AI coding inference costs by roughly 90 percent in under a year by making self-hosted models its default instead of commercial services like Claude. Ops and infrastructure lead David Abitbol said coding time improved 60 percent for roughly 2,000 engineers. Running GLM 5.1 at $0.30 per million input/output tokens cost about a tenth of OpenAI's Codex or Anthropic's Claude Code, Abitbol said.
Latest updates
Revolut Cuts AI Coding Inference Costs by About 90% Using Self-Hosted Models
Revolut says it has cut its AI coding inference costs by roughly 90 percent in less than a year. The company achieved this by making self-hosted models its default rather than relying primarily on commercial services such as Claude. Revolut made the claim while speaking at the AI Engineer conference in New York, according to The Register. The company said cheaper tokens have led engineers to use AI for time-saving tasks such as debugging and working through Jira tickets. The figures are the company's own and have not been independently verified.