Terray says TerraBind delivers 26x faster inference than Boltz-2 by skipping diffusion-based coordinate generation

When MIT and Recursion released Boltz-2 last June, it marked a milestone: the first open-source AI model to jointly predict protein structure and small-molecule binding affinity. The model approaches the accuracy of physics-based free-energy perturbation methods at roughly 1,000x the speed. Its predecessor Boltz-1 had already been widely adopted, and Boltz-2 built on that momentum.… The post Terray says TerraBind delivers 26x faster inference than Boltz-2 by skipping diffusion-based coordinate generation appeared first on Drug Discovery and Development.

Continue reading →