British fintech Revolut has significantly reduced its AI coding expenses by adopting a self-hosted model, achieving a 10x cost reduction. The company transitioned away from using proprietary models like Claude for general coding tasks, instead opting for open-source alternatives such as GLM. While Claude and Gemini were previously experimented with, Revolut found that the primary bottleneck was not the AI model itself, but rather its application to tasks like debugging and managing Jira tickets. The company still utilizes models like Claude for vision-specific work, as GLM is text-only. AI
IMPACT Suggests a trend towards self-hosted and open-source models for cost efficiency in enterprise AI development.
RANK_REASON Article discusses a company's internal adoption of AI tools and cost-saving measures, rather than a new release or significant industry-wide event.
AI-generated summary · Google Gemini · from 5 sources. How we write summaries →