Revolut moves AI coding to self-hosted GLM models
Revolut has made self-hosted Z.ai GLM-5.3 its default AI coding model for about 2,000 engineers, running on 56 Nvidia B200 GPUs, according to TechRadar, 36氪 and The Next Web. The company reports that inference costs fell about 90%, coding time fell 60% and lead time improved 70%, while commercial models remain in use for visual and specialized tasks. Revolut says it initially imposed token limits after some engineers spent $1,000 to $10,000 per month without clear returns.
The deployment provides a detailed enterprise example of combining self-hosted open-weight models with commercial frontier models, but its productivity figures are internal and do not isolate the model change from other factors.
有人每个月花 1 万美元用 AI 编程，结果老板一看账单：这钱花得值吗？——这正是英国金融科技公司 Revolut 曾经遇到的尴尬。
为了让工程师用上 AI，公司先后尝试了 Gemini、Claude Code，但部分工程师每月就能烧掉 1000 美元甚至 1 万美元。可钱花出去了，效率提升却没有达到预期。于是，Revolut 做了一个大胆的决定：不再把 AI 编程的主力押在商业模型上，转而自己去部署智谱的开源模型 GLM。
结果相当惊人：不到一年，AI 编程推理成本下降约 90%，公司约 2000 名工程师的编码时间缩短了 60%，从接手需求到代码合并的交付周期也缩短了约 70%！
Claude 太贵了！有工程师每月烧掉 1 万美元
一开始，Revolut 其实并没有打算自己训练或部署大模型。
2025 年初，公司开始用 Gemini 帮工程师查找资料、阅读技术文档；到了 2026 年初，又引入了 Claude Code，希望进一步提高开发效率。可随着使用规模扩大，成本很快成了问题。
#Revolut#Z.ai#GLM-5.1#GLM-5.3#Nvidia B200#Claude Code#Gemini#GPT#Codex