# Phikon / Phikon-v2 (Owkin)

Source: https://onco.cc/technologies/phikon/  
OnCo record `phikon` (Technology). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

Owkin's open pathology models trained on TCGA and its federated hospital network.

## Summary

Phikon is Owkin's line of open pathology foundation models, vision transformers trained with iBOT and DINOv2 self-supervision. Phikon (2023) was trained on TCGA, and Phikon-v2 (2024, 460M tiles from 58M slides of 30 cancer types) drew on Owkin's federated hospital network, in which models learn from data that stays inside partner hospitals. Both power Owkin's MSI and prognosis products, and the open weights are used by academic groups as backbones for their own classifiers. Compared with the largest pathology models Phikon is of moderate scale, and its federated data access is an advantage for training but harder for outsiders to audit. For a newcomer: Phikon is an openly released pathology model from a company that trains on hospital data without moving it.

## Fields

- Kind: Technology
- Status: emerging
- Last checked: 2026-09-08
- Tags: foundation-model; pathology
- Principle: Phikon is trained with iBOT/DINOv2 self-supervision.
- Since: 2023
- Strengths: Open weights; Federated data access
- Limitations: Moderate scale

## Sources

- Phikon-v2 (arXiv 2024): https://arxiv.org/abs/2409.09173

## Connected records

- fronts: [AI & Computation](https://onco.cc/fronts/ai-computation/)
- technologies: [Pathology & radiology foundation models](https://onco.cc/technologies/pathology-foundation-model/)
- companies: [Owkin](https://onco.cc/companies/owkin/)

---
JSON: https://onco.cc/api/v1/entities/phikon.json