Radical AI operates an autonomous closed-loop platform for materials discovery. The platform screens billions of material compositions to predict structures and physical properties and identify experimental candidates. It optimizes chemical synthesis through computational adaptive experimentation, active learning, and self-guided review of scientific literature. High-throughput experiments on selected candidates occur in a self-driving laboratory. Data generated feeds back into the prediction engine to improve the process. The company targets mission-critical industries requiring breakthrough materials beyond incremental progress. Radical AI released LitXBench, a benchmark evaluating large language models on extracting experiments from materials science literature.