As AI systems become increasingly integrated into daily life, their capacity to both enhance and undermine human flourishing demands rigorous assessment beyond narrow technical performance metrics. This position paper argues that interactive benchmarks that measure the quality of human-AI behaviors are essential for developing AI that genuinely supports human flourishing. We propose a framework for designing and implementing interactive benchmarks for various dimensions of human flourishing and discuss methodological challenges in capturing the complex, longitudinal nature of human-AI interactions. Drawing from multidisciplinary research on well-being, we identify six critical domains where such benchmarks are needed to evaluate the potential for AI systems to encourage flourishing and the risk of negative outcomes. This work establishes a foundation for AI development practices that prioritize human flourishing as a central objective rather than an incidental outcome.