Guidepoint Labs
How frontier models perform in clinical trial design
2 expert-graded benchmarks covering early phase and late phase lung cancer trials


Get the white paper
Free. We’ll email you a copy of it.
WHAT’S INSIDE
What 468 expert evaluations found
Results from two Guidepoint Labs benchmarks that tested frontier AI models on clinical trial design, graded blind by vetted Guidepoint Advisors.

How close models get to expert-grade
Scores for eight frontier and specialist models on an anchored 1-to-5 scale, and how many responses reached the expert-grade standard.

Where models are least reliable
Why source discipline ranked last for every model in both benchmarks, and what that means for clinical AI workflows.

What an automated grader would miss
Examples from our Advisors’ written rationales of confident, well-formatted errors that only a specialist would catch.

How the benchmarks were built
The rubric, Advisor certification, blind three-read design, and full review behind every score in the report.
WHY GUIDEPOINT
Not all intelligence
is equal.
Guidepoint connects organizations with the human expertise needed to make AI more relevant, reliable, and useful.