Google published results from a foundation model family (Cell2Sentence‑Scale/Gemma) that proposed and helped validate a new cancer drug combination, marking a milestone in AI‑driven biomedical discovery. The model—trained at large scale—generated hypotheses that were taken into preclinical validation pipelines, where indicated synergistic combinations exhibited activity, according to the company and coverage in BioTecNika. The announcement highlights rapid progress in applying large‑language and multimodal models to drug discovery tasks, from target identification to combination prediction. Researchers caution that such results require independent replication and regulatory pathways for AI‑generated candidates remain nascent, but the work signals growing industry uptake of large AI architectures to accelerate early‑stage discovery.