Personal Agent Bench

Personal Agent Bench is an independent benchmark for personal AI agents, the assistants that book, cancel and schedule things for you. We take each product apart at the runtime level (where the model runs, how tools reach it, where memory lives, what you approve) and run it through a repeatable task exam. Every completion rate comes with a 95% interval and its sample size; there is no combined score and no ranking. Run by the Zentor team, which builds agent software, so Zentor's products stay off the scorecard.