Insilico Medicine launched what it calls the first “Drug Discovery and Development Benchmark-as-a-Service” aimed at testing frontier AI models for decision-making rather than memorizing training data. The framework is designed to measure whether AI systems can make meaningful scientific choices. The company’s announcement targets a widely discussed gap in AI evaluation: many existing benchmarks do not reliably distinguish robust reasoning from recall performance on training sets. Insilico’s offering is framed as a standardized way for stakeholders to assess scientific capability. For the biotech sector, the release is another attempt to move AI from concept to audited toolsets that can support target discovery, molecule design, and development planning. The next test will be adoption—whether sponsors, research groups and technology partners integrate the benchmark output into go/no-go decisions for AI-assisted R&D programs.