In building Gina—our AI-driven crypto wallet—we evaluated over six different models on real user tasks, from market price checks to multi-step token bridging. This post outlines the methods we used to compare speed, accuracy, transaction success rates, and user experience. If you’re exploring how to integrate AI and on-chain tooling, our findings offer a clear view of each model’s strengths, shortcomings, and overall suitability for agentic crypto tasks.