Learn about this first
According to David Abitbol, Revolut’s head of AI operations, Revolut has moved approximately 2,000 engineers to self-hosted GLM models for coding, while placing Claude Code and Codex on an exception list.
Abitbol said the setup uses 56 Nvidia B200s, with a blended cost of 30 cents per million tokens—approximately one-tenth of what the company previously paid Anthropic and OpenAI; the engineering delivery cycle has been shortened by 70%. These cost and efficiency figures are claims from Revolut.
Karina Nguyen reposted the news and commented that every company should turn itself into a giant RL environment, building self-improving AI that learns from every interaction and gradually takes over company operations.
AI-generated from reports · updated 13 days agoLatest developments
Karina Nguyen says, based on Revolut's migration case, that enterprises should turn themselves into giant RL environments
Popularity trend
Event progress1
Revolut migrates to self-hosted GLM models
Hacker News: Hot AI topic: Revolut moves approximately 2,000 engineers to self-hosted GLM for coding