Researchers from MIT and partners released Boltz‑2 and highlighted the rise of physics‑grounded large quantitative models (LQMs) for drug discovery. Boltz‑2 led binding‑affinity prediction at CASP16, producing affinity estimates in ~20 seconds—orders of magnitude faster than free‑energy perturbation methods—and the code was released under a permissive MIT license. The work suggests AI models are moving beyond structure prediction toward rapid affinity ranking and could accelerate hit‑to‑lead workflows when combined with proprietary datasets and experimental validation.