Bioptimus

#Overview

Bioptimus is a pathology foundation-model lab and the first-party source of H-optimus-0, a foundation model for digital and computational pathology. Researchers searching for histopathology embeddings or a whole-slide-image feature extractor can access H-optimus through Bioptimus's API. Pathology and medical imaging are blind spots for the protein-focused platforms, which makes first-party labs like Bioptimus the natural route to these models.

#What you can run on Bioptimus

• H-optimus-0 — a 1.1B-parameter vision transformer for histopathology, trained on 500,000+ H&E whole-slide images from thousands of clinical practices. It produces general-purpose embeddings from tissue image tiles, supporting downstream tasks such as tumor detection, tissue classification, biomarker prediction, and slide retrieval.

This covers the computational pathology and digital histopathology domain. The open H-optimus-0 weights are available on Hugging Face, and Bioptimus provides hosted API access.

#API access on Bioptimus

Bioptimus offers hosted inference for H-optimus through its API, letting pathology and biomedical imaging teams extract slide-level and tile-level embeddings at scale without hosting the model themselves. This is aimed at computational pathologists, biomarker and diagnostics researchers, and machine-learning teams building weakly-supervised whole-slide-image pipelines. The open H-optimus-0 weights also remain on Hugging Face for self-hosted inference when you would rather run the encoder inside your own environment.