Performance results of AI coding models on Speakeasy tasks, measuring success rate and execution time with high precision.