Same sourcing standard as the off-the-shelf dataset — every fact cited, nothing estimated — scoped to your industry, your data shape, and your volume.
Custom eval build — from $2,500
Tell us your domain and what your AI must not get wrong. We build a ~50-row verified eval for your exact use case in about a week: every row independently derived from live primary sources, cited, honestly verdicted, and shipped in the machine-runnable eval format with a graded frontier-model baseline. Larger scopes quoted individually.
Hallucination Report — from $999
We run our verified benchmark against your model (your API endpoint, or you run our runner and send the raw outputs) and deliver a signed report: hallucination rate per domain, every failure with the primary-source evidence, and a methodology appendix — documentation you can hand to customers, auditors, or compliance reviewers asking for accuracy evidence.
Continuous drift monitoring — quoted
For products that assert facts: we re-verify your claim corpus against primary sources on a monthly cadence and report exactly what drifted, with dated source evidence. Same discipline our own Living datasets run on.
Use the form below for any of these — say which one you're after in the details field.