Google’s Cell2Sentence‑Scale foundation model (27B parameters) reportedly proposed and validated a cancer drug combination, marking an AI‑driven step in hypothesis generation for oncology. Meanwhile, academic and commercial groups released fast, physics‑aware models (e.g., Boltz‑2) that predict binding affinity and accelerate small‑molecule discovery workflows. Together these developments lower barriers to in silico lead identification and prioritization. Companies and academic labs are adopting large quantitative models and open‑source tools to shorten early discovery cycles, but experts note that integration with experimental validation and careful benchmarking remain essential before translating model outputs into clinic‑ready candidates.