# Revolut moves to self\-hosted GLM

- HackerByte page: <https://hackerbyte.net/en/stories/fintech-giant-revolut-s-steps-towards-self-hosting-large-language-models-llm-50350004>
- Original source: <https://twitter.com/guilleflorvs/status/2117379073785696395>
- Hacker News discussion: <https://news.ycombinator.com/item?id=50350004>
- Submitted to Hacker News: 2026-11-03T09:53:59.000Z

Fintech giant Revolut's steps towards self\-hosting large language models \(LLM\).

## AI-generated summary

Revolut has transitioned approximately 2,000 engineers to a self\-hosted GLM for coding tasks, removing Claude Code and Codex from general use. This move was driven by cost savings, with a blended cost of $0.30 per million tokens, significantly lower than previous API costs. The change also led to a 70% improvement in engineering lead time as engineers utilized the tool more broadly.

## Key points

- Revolut moved 2,000 engineers to a self\-hosted GLM.
- Claude Code and Codex are now exceptions, not the default.
- Self\-hosted GLM costs $0.30 per million tokens.
- Engineering lead time improved by 70%.
- Open\-weight models are becoming the default for Revolut.