A benchmarking framework for organoid computing

Organoid computing leverages functional neural networks of lab-grown neural organoids to recapitulate key aspects of human brain information processing, holding promising potential in brain-inspired computing, brain-computer interfaces, neuroscience, and medicine. However, its broad application remains constrained due to technical limitations including biological variability, interface instability, stochastic performance, and limited reproducibility. Here, we present a benchmarking framework that systematically quantifies the key performance determinants of organoid computing. We operationally define ‘benchmark-ready’ organoid substrates using optional criteria encompassing viability, cellular identity, structural integrity, functional maturation, and baseline networks, supported by multimodal phenotyping. We further establish qualification metrics for organoid-machine interface and intrinsic neural and network function, including physical coupling, spatial coverage, temporal stability, spontaneous activity, and network-level organization. Finally, we introduce a hierarchical evaluation framework for stimulation-evoked computation combined with information encoding criteria, standardized scoring, and statistical reporting. We believe that the benchmarking framework may provide an adaptable foundation for standardized, reproducible and comparable evaluation of organoid computing technologies, thereby facilitating the broad adoption and application of organoid intelligence.