Benchmarks.Bio partnered with Fireworks AI to add its BioBench suite to Fireworks’ Specialized Intelligence Index (SII), elevating real-world agentic biology evaluation from research demos into benchmark infrastructure. BioBench covers 1,337 life-science tasks, designed to mimic end-to-end laboratory and discovery workflows with 50+ turns per task and runtimes up to 10 hours. The benchmark spans omics workflows, therapeutics testing for modalities such as ASOs, siRNAs, and antibodies, clinically relevant application tasks, and refusal grading for disguised dual-use biology red-team scenarios. By placing the benchmark into an index used to measure agent capability on production-grade tasks, the partnership targets an industry need: reliable evaluation under realistic tool-and-instrument constraints. For biotech organizations evaluating AI copilots and agent systems for R&D pipelines, this is a practical shift toward standardized, safeguards-aware performance testing across the kits and providers encountered in day-to-day work.