TITAN

A multimodal whole-slide foundation model for computational pathology from the Mahmood Lab (Nature Medicine 2025), pretrained with visual self-supervision and vision-language alignment on 335K whole-slide images. TITAN produces slide-level representations for zero/few-shot classification, retrieval, and pathology report generation, and is one of the field's most-adopted slide FMs (~150K HuggingFace downloads). Pretrained from scratch on pathology data by Harvard/Mass General Brigham.