A new transparent benchmark focused on evaluating the actual capabilities of AI coding agents has been compiled by MyToken. According to PANews, this benchmark assesses success rates as the primary dimension, while speed and cost are considered separate dimensions for future analysis. The benchmark is fully open and reproducible, presenting rigorous evaluation standards along with the latest top 10 rankings based on success rates.