Mithrl raised $20 million in a Series A round led by Obvious Ventures, expanding development of its biomedical world model and agentic AI “harness” for model routing, token optimization, and content orchestration. The company said the platform is built to help R&D teams move from idea to IND faster by organizing validated biology and integrating model outputs with proprietary and partnered datasets. Mithrl positions its approach as reasoning over curated, peer-reviewed biology rather than exploring broader unverified information spaces, aiming to reduce compute needs and improve traceability of system outputs. The company also launched an early access program for its second-generation platform, Mithrl-1, beginning Sept. 21. The company reported benchmark results claiming a 45% reduction in tokens versus standard workflows running the same frontier models without customization. Mithrl also said hypotheses include the sources used plus a confidence score designed to reflect trustworthiness and reading provenance. The round signals continuing venture demand for AI infrastructure that can connect evidence and execution planning for regulated drug development workflows.