Microsoft releases GigaPath-Flash and GigaTIME-Flash to reduce computational costs in pathology foundation models

Microsoft Research released GigaPath-Flash and GigaTIME-Flash, two efficient pathology foundation models designed to reduce computational demands while maintaining competitive performance on whole-slide benchmarks.

GigaPath-Flash combines a 22M-parameter ViT-S tile encoder with a 21M-parameter LongNet slide encoder, both distilled from the original billion-parameter GigaPath encoder. The tile encoder transfers representational capacity from the larger model into a backbone an order of magnitude smaller.

On slide-level classification benchmarks such as PANDA prostate grading and EBRAINS brain tumor subtyping, GigaPath-Flash achieves the lowest inference cost among whole-slide pretrained models while scoring within 3% of the original GigaPath at roughly 50 times less compute.

GigaTIME-Flash replaces the CNN backbone of the original GigaTIME with the GigaPath-Flash ViT-S encoder, paired with a lightweight convolutional decoder for H&E-to-mIF translation. The model is fine-tuned using LoRA adapters, keeping the pretrained encoder weights largely frozen.

On both in-distribution and out-of-distribution cohorts spanning brain, breast, colon, and lung cancers, GigaTIME-Flash matches or improves upon the original GigaTIME in spatial protein prediction quality. The gains are particularly notable on out-of-distribution data, suggesting improved generalization.

The Flash family models are released under the Apache 2.0 license and are positioned as research models. Microsoft explicitly states they are not intended or validated for clinical use, including diagnosis, prognosis, or treatment selection.

The researchers highlight that whole-slide images often exceed a gigapixel, and applying foundation models to even a single slide requires processing thousands of image tiles. Computational cost limits the number of patients, datasets, tasks, and hypotheses that can be studied, making efficiency improvements critical for population-scale discovery.