Revolut says self-hosted GLM cut its AI coding costs by 90% for 2,000 engineers
David Abitbol says 56 Nvidia B200 GPUs serve GLM to about 2,000 engineers at $0.30 per million tokens, with Claude Code and Codex kept for exceptions.
By Ryan Merket · Published
Primary source: news.ycombinator.com
Why it matters
Revolut's account separates model quality from usage economics: the reported productivity gain arrived when engineers stopped rationing tokens. For founders and engineering leaders, the case shows when self-hosting can change behavior at scale, and it also shows the entry price - 56 B200 GPUs, serving expertise and sandbox controls.
Revolut has made self-hosted GLM its default coding model for about 2,000 engineers, pushing Anthropic's Claude Code and OpenAI's Codex into exception roles. David Abitbol, who leads AI operations at Revolut, detailed the deployment at AI Engineer NYC on October 13th, and Guillermo Flor recirculated the figures in a post on November 2nd. Revolut says self-hosting began around April 2026.…
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.