Skip to content

Ratings Tasks Compare Releases How we test

First release in progress

Independent tests of AI models on real product management work. Graded by Jev, reviewed blind by Martin.

Compare configurations

Pick up to three. The URL is the comparison, so share it.

Configuration 1Configuration 2Configuration 3

Choose at least one configuration.

PM Benchmark measures performance on this suite of tasks under these configurations. It is not a universal measure of product management ability.

Structured grading by Jev, TypeSafe’s decision model, treated as a grader under calibration. Blind human review by Martin Slaney.

The benchmark

Ratings
Task library
Compare configurations

The method

How we test
Scoring formula
Release history