Dragon (LUMC/EU)

Moorfields/UCL retinal foundation model (MAE pretrained on 1.6M retinal images) for disease detection across ophthalmic and systemic conditions.

Google open medical foundation model (Gemma-based) with text and multimodal medical-image variants released under Gemma terms.

Multimodal radiology foundation model handling 2D/3D scans + text; released with MedMD dataset for generalist radiology AI research.

Vision-language pathology foundation model trained on 1.17M histopathology image-caption pairs — enables zero-shot CPath.

Saama open-weights biomedical LLM fine-tuned on Llama-3 70B; competitive with closed clinical models on MedQA/MMLU-Med.

Mahmood Lab general-purpose pathology foundation model trained on 100M+ histology image tiles; SOTA across CPath benchmarks.