This JevBench is a decision workbook.

jevbench.app helps one person get one Jev decision working. You bring a question and examples you label. JevBench shows where Jev’s answer differs from those labels, lets you explore the tradeoff on that sample, and lets you revise the question and check a fresh set.

The first 10 examples of one decision are a free early look. $9, once, continues that same decision on the web, through the API, or from a coding agent. A small sample is not a production cutoff.

Other public projects use the same name

This site is not the JevBench leaderboard published by Benchmark Heaven. That project is at benchmarkheaven.com, with code at github.com/fstandhartinger/jevbench.

This site is also not jevbench.xyz. That is a separate public project.

Those are different efforts. This page only says which one you are on.

What is Jev? · Playground · One recorded experiment · Pricing