Revolut cuts AI coding costs 10x by self-hosting GLM-5.3 instead of Claude
Revolut reduced its AI coding inference costs by about 90% in under a year by shifting from commercial services like Anthropic’s Claude Code to self-hosted models. The fintech now uses GLM-5.3, powered by 56 Nvidia B200 GPUs, after initially testing GLM-5.1 and earlier models like Gemini and Claude Code. Engineers’ monthly token spend dropped from thousands of dollars to pennies, freeing them to…
Key points
- Revolut slashed AI coding costs by ~90% in under a year by self-hosting GLM-5.3 instead of Claude Code
- Engineers’ coding time improved by 60% and lead time by 70% with cheaper, unlimited token access
- GLM-5.3 runs on 56 Nvidia B200 GPUs, optimized for 256K-token context and 200 tokens/sec throughput
David Abitbol, Revolut’s ops and infrastructure lead, reported a 60% speedup in coding and a 70% reduction in lead time for its 2,000 engineers. Throughput of 200 tokens/second meets engineers’ needs, though after-hours users hit 800 tokens/second. GLM handles text-only tasks; vision queries still route to models like Gemini or GPT Luna.
Model pages: GLM 5.3 → · Claude Code →
British fintech Revolut slashes AI coding costs 10x by cutting out Claude and going self-hosted
theregister.com · 3 November 2026
Loading the full article…
This text was published by theregister.com and written by Tobias Mann. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
1source- Hacker News discussion · 9 pointsnews.ycombinator.com
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Hardware & Compute
All →- WhiteFiber buys two North Carolina sites for $60 million · 1 src
- Hermantown approves Google’s 1.7M-sq-ft Minnesota AI data center after concessions · 1 src
- Innolight’s first near-package optics device begins early production with Tower Semiconductor · 1 src
- Supermicro and SambaNova team on SN50-powered AI inference systems for cloud and enterprise · 1 src
- Lumai deploys Iris optical servers in ARIA’s £50M Scaling Inference Lab · 1 src
Comments
via GitHub Discussions