Pricing
Priced per deployment, not per patient
HLA-Verify is in private preview, so pricing is set per pilot rather than published as a rate card. What follows is the shape of the tiers and what separates them. Early pilots get preferential terms and a direct line to the person who built the engine.
Evaluate
Free
- Interactive demo and
/v1/verifyagainst the current release - Rate-limited, current release only
- Full HLA-Bench-A results and dev split
- Source readable; noncommercial use permitted
- No key, no contract, no data retained
Laboratory
Per lab, annual · most pilots start here
- All three endpoints with an issued API key
- Pinned release of your choosing — it moves when you say so
- Batch
/v1/normalizefor LIMS exports and historical harmonization - Audit log of call counts per key (counts, never content)
- Uptime commitment and a named contact
- Self-hosted container option for closed networks
OEM & Enterprise
LIMS, diagnostics and AI product vendors
- Embed verification inside your product
- Commercial licence for the service code
- Multi-release pinning and release-diff reporting
- Private deployment, your infrastructure or ours
- Support terms your customers can be shown
Evaluation & environments
AI labs and data vendors · project-based
- Sealed-split evaluation of your model, versioned report
- Custom task families with deterministic graders
- Export to
verifiers, Inspect AI or Harbor - Regeneration subscription — new IPD release, new suite
What is the same at every tier
- Nothing is stored. Not on the free tier, not on enterprise. Request content is processed in memory and discarded; we log counts, not content.
- No LLM in the loop. There is no cheaper tier that swaps determinism for a model. That would be a different product.
- The same engine. The free demo and the OEM deployment run identical verification code against the same reference data. The tiers differ in pinning, limits, support and licence — never in correctness.
- Reproducible answers. Any verdict can be recomputed from the release it names, for as long as that release exists.
Questions we expect
Can we run it inside our network?
Yes. The service is a container with two environment variables. Reference files are fetched from the official source at startup and MD5-verified; after that nothing needs to leave. This is the usual answer for labs whose answer to "can data leave" is no.
What happens when a new release lands?
Nothing, until you decide. Your key stays pinned to the release you chose. We publish what changed between releases; moving is an explicit action with a diff you can review first.
Is this a medical device?
No, and it is not marketed as one. It is a research-and-evaluation tool that checks names against a reference database. It supports your validated procedures; it does not replace them or make clinical determinations. See Trust.
Why not just build this in-house?
You can — the source is readable and the approach is not secret. What takes the time is the resolution rules at the edges: ambiguous G groups, deleted names without successors, prefix validity, expression suffixes, serology's meaningful zero. Those are the cases the benchmark is built out of.