Mecha Health

Mecha Health builds foundation models for radiology. The premise is direct: medical images go in and complete radiology reports come out. Rather than flagging a single finding, the models are trained to interpret an image and generate the full narrative report a radiologist would write, which positions the product as a drafting and reading assistant for high volume imaging work. It is intended for radiology groups, hospitals and imaging providers looking at report automation. The company's published research centers on chest X ray interpretation, including its Mecha Net models for X ray report generation, methods for evaluating automated radiology reports, qualitative case studies for assessing foundation models, and analysis of why radiologists make mistakes. The most recent model release described on the site, Mecha Net v0.2, is presented as reaching human performance on its report generation task. The company is based in San Francisco, went through Y Combinator's Winter 2025 batch and announced a $4.1 million seed round in November 2025. Pricing is not published. There is no self serve product or plan page; prospective customers book a call with a founder to discuss deployment. Buyers should treat this as an early stage clinical AI company. The site does not state regulatory clearance, supported modalities beyond chest X ray, or integration details for PACS and reporting systems, so those questions need to be answered directly with the company before any clinical use.