An MIT research team released BoltzGen, an open‑source generative model that moves from structural prediction toward generalizable therapeutic design across modalities including small molecules, nanobodies, and peptides. The model is available under an MIT license and collaborators report initial wet‑lab validations showing nanomolar affinities for diverse targets. BoltzGen builds on prior Boltz models, aiming to design binders against ordered and disordered proteins and to propose solutions for previously undruggable targets. The authors stressed emphasis on unsolved problems; a network of 26 academic and industry partners is running validation experiments across antimicrobial and oncology applications. The permissive licensing and early validation could accelerate internal drug discovery at startups and pharma, while raising intellectual‑property and commercialization considerations for companies integrating open models with proprietary data.