{"entity":{"id":"uni-conch","kind":"technology","name":"UNI and CONCH (Harvard, Mahmood Lab)","aka":[],"tldr":"Two open academic pathology models: UNI reads tissue images, CONCH links images with pathology text.","summary":"UNI and CONCH are two open academic pathology foundation models from the Mahmood Lab at Harvard. UNI is a vision encoder trained with DINOv2 self-supervision on 100 million tiles from 100,000 slides (Nature Medicine 2024), with the larger UNI2-h scaling it further. CONCH (Nature Medicine 2024) is a vision-language model trained on 1.17 million image-caption pairs by contrastive alignment, which enables zero-shot classification and image or text retrieval without task-specific training. Both are released with open weights under a non-commercial licence and are widely used as reproducible baselines for cancer subtyping, biomarker prediction and prognosis. They work at tile level, so slide-level decisions need an aggregation step, and clinical validation is task by task. For a newcomer, UNI learns what tissue looks like and CONCH learns what pathologists say about it.","status":"emerging","asOf":"2026-09-08","links":[{"label":"UNI, Nature Medicine 2024","url":"https://doi.org/10.1038/s41591-024-02857-3"},{"label":"CONCH, Nature Medicine 2024","url":"https://doi.org/10.1038/s41591-024-02856-4"}],"tags":["foundation-model","pathology"],"related":[],"cancers":[],"sections":["ai-computation"],"technologies":["pathology-foundation-model"],"targets":[],"drugs":[],"companies":[],"institutions":["dana-farber"],"pathways":[],"terms":[],"trials":[],"people":[],"bottlenecks":[],"keyPapers":["paper-chen-nat-med","paper-lu-nat-med"],"journals":[],"dependsOn":[],"notes":[],"principle":"UNI is trained with DINOv2 self-supervision; CONCH uses contrastive image-text alignment.","strengths":["Open weights for research","Reproducible baselines"],"limitations":["Non-commercial licence","Tile-level, needs aggregation"],"since":2024},"route":"/technologies/uni-conch/","neighbours":{"section":[{"id":"ai-computation","kind":"section","name":"AI & Computation","route":"/fronts/ai-computation/"}],"technology":[{"id":"pathology-foundation-model","kind":"technology","name":"Pathology & radiology foundation models","route":"/technologies/pathology-foundation-model/"},{"id":"titan","kind":"technology","name":"TITAN (whole-slide multimodal model)","route":"/technologies/titan/"}],"institution":[{"id":"dana-farber","kind":"institution","name":"Dana-Farber Brigham Cancer Center","route":"/institutions/dana-farber/"}],"paper":[{"id":"paper-lu-nat-med","kind":"paper","name":"A visual-language foundation model for computational pathology","route":"/key-papers/paper-lu-nat-med/"},{"id":"paper-chen-nat-med","kind":"paper","name":"Towards a general-purpose foundation model for computational pathology","route":"/key-papers/paper-chen-nat-med/"}],"roadmap":[{"id":"ai-oncology-roadmap","kind":"roadmap","name":"AI in oncology roadmap: pattern readers → foundation models → agents in the workflow","route":"/roadmaps/ai-oncology-roadmap/"}]}}