Coinbase has successfully reduced its AI spending by nearly 50% by implementing open-weight models and optimizing its infrastructure, according to CEO Brian Armstrong. The company has shifted to using models like Zhipu's GLM 5.2 and Moonshot AI's Kimi 2.7 as default options, allowing engineers to select the most appropriate model for specific tasks. This strategic move has enabled 91% of employees to avoid hitting usage caps, thus reducing costs without limiting AI usage.
Coinbase has also improved its routing and caching strategies. By preprocessing prompts and routing tasks to the most suitable models based on cache hit rates and pricing, the company has increased its cache hit rate from 5% to 60%. Armstrong emphasized that the goal is to create a sustainable infrastructure that supports exponential growth, focusing on reducing wasted tokens and enhancing usage visibility to ensure high-impact outputs from AI investments.
Coinbase Halves AI Spending by Adopting Open-Weight Models
Disclaimer: The content provided on Phemex News is for informational purposes only. We do not guarantee the quality, accuracy, or completeness of the information sourced from third-party articles. The content on this page does not constitute financial or investment advice. We strongly encourage you to conduct you own research and consult with a qualified financial advisor before making any investment decisions.
