The radiology foundation model. Developer-ready.
Build AI solutions for imaging workflows, from discovery to deployment in weeks, not years.
Radiology AI shouldn’t require starting from zero.
Building imaging AI means assembling training data, developing models from scratch, and validating across diverse clinical environments — and that’s only part of what it takes. It takes years and millions of dollars, if you can do it at all.
Start with a foundation model built for radiology.
Harrison.Rad.1.5 is a foundation model purpose-built for radiology. Unlike a GPT, it gives software developers a solution starting point for building AI across operational, research, and clinical use cases.
See Harrison.Rad.1.5 in action
Upload X-rays and test use cases firsthand on chat.harrison.ai
You build and own the solution.
Unlike black-box solutions, you own the IP you developed for your own use case.
From kickoff to go-live in weeks, not years.
Define use cases. Get API access.
Run Harrison.Rad.1.5 against your historical cases. Measure quality before any workflow changes.
Your data + our training cases.
Limited deployment.
Continuous improvement. Ongoing model updates.
Questions we hear most.
A foundation model is a single large-scale AI model that can be adapted to many different tasks. Harrison.Rad.1.5 was built from the ground up on radiology data, validated by radiologists, and designed to power whatever workflows you choose to build.
General-purpose models weren’t trained for radiology. Rad.1.5 understands radiological language, image patterns, and clinical workflows in ways generic models simply don’t.
That’s one of the use cases our partners are exploring. With your own prompts and templates, harrison.rad-1.5 can draft radiology reports. This is not an approved or cleared medical device. Regulatory approval, validation, and compliance for your use case are your responsibility, where required.
Some partners are exploring using the model to route cases to specific worklists based on one or more criteria you define.
It’s a use case partners are investigating. The model can check agreement between radiology and procedure codes. How you build on top of this is up to you.
A few months, not years. We start with a quick Discovery to define your use cases, run a retrospective review against your historical data, develop with your prompts and training cases, then move to a controlled launch to resources, then full deployment with ongoing monitoring.
Yes. Our BAA is HIPAA-compliant and we take data security seriously across every deployment.
A point solution does one thing. A foundation model is infrastructure you build on. One model our partners can own, delivering value exactly tailored to your workflow — instead of buying someone else’s product.
A conversation. We start with a quick discovery to define your use cases and review validation data.
This is not an approved or cleared medical device, and harrison.rad-1.5 is not FDA cleared. Regulatory approval, validation, and compliance for your use case are your responsibility, where required.
Ready to build?
Software developers can instantly start your Harrison foundation model based AI used cases.