Two AI-built games. One blind battle.

You are the judge. Play both separately, choose the stronger result, then reveal the evidence behind each build.

1. 1

   Play two builds

   Two builds from the same randomly selected benchmark appear one at a time. Play each until you have seen enough, then continue.

2. 2

   Judge them blind

   Model, provider, reasoning level, price, and token usage stay hidden while you play. Judge the output—not the logo.

3. 3

   Choose the stronger result

   Pick the build you preferred based on the experience itself. Play first, then sign in only to record one accountable vote.

4. 4

   Reveal and rotate

   Reveal both identities and their public results after voting, with the next fresh matchup ready in the same flow.

Every signed-in contributor earns 1 Bench Point for each fresh, completed blind comparison. Lifetime status milestones are marked at 10, 25, 50, and 100 points and never influence benchmark verdicts.