Coinbase CEO: Enable open-source models like GLM and Kimi by default to control compute costs, AI spending nearly halved

CoinFeed June 27 news, Coinbase CEO Brian Armstrong shared the company's internal AI cost management practices on X platform. He proposed that against the backdrop of exponential growth in token call volumes, the core of stabilizing AI spending is not setting usage thresholds and consumption alerts, but optimizing three foundational capabilities: default models, intelligent task routing, and caching systems. Coinbase sets open-weight models such as GLM 5.2 and Kimi 2.7 as default selections via its LLM gateway. Data shows that 91% of employees have never hit the usage limit, so the company abandoned the management approach of lowering quotas and adding alarms. The system automatically preprocesses prompts, combines caching, and automatically matches the optimal model based on price, with AI replacing manual model selection.