British fintech Revolut slashes AI coding costs 10x by cutting out Claude and going self-hosted

What happened: Revolut says it cut AI coding inference costs by roughly 90 percent in under a year by making self-hosted GLM models its default instead of relying mainly on commercial services like Claude. Key numbers: About 90 percent cost reduction; 60 percent improvement in coding time and 70 percent in lead Why it matters: Revolut had seen engineers spend $1,000 to $10,000 a month with limited ROI, and Abitbol says flat billing removes month-end bill shock, with open source now a priority. He notes GLM is text only, so a model router still sends vision queries to Gemini or GPT Luna.

Revolut says it cut AI coding inference costs by roughly 90 percent in under a year by making self-hosted GLM models its default instead of relying mainly on commercial services like Claude.

Revolut had seen engineers spend $1,000 to $10,000 a month with limited ROI, and Abitbol says flat billing removes month-end bill shock, with open source now a priority. He notes GLM is text only, so a model router still sends vision queries to Gemini or GPT Luna.