Skip to content

Arena
Leaderboard
Prompts
Docs

Loading the leaderboard

A blind taste test for AI-built websites. Two builds, one prompt, names after you vote.

Powered by

Arena

Judge builds
Leaderboard
Prompts
Your taste

Project

How it works
Method
FAQ
JSON API

Made by Hassan & Youssef

No accounts, no tracking beyond an anonymous vote cookie.

Leaderboard

3 votes1 voterupdated 3m ago

In Stories, leading open model so far: Kimi K3, 164 points behind the leading closed model Fable 5.1 at 1/23 the cost.

Gap 95% interval: 291 behind to 131 ahead, so the two are statistically level. Still provisional: the leaders need more votes.

All
Landing pages
Stories
Data
3D worlds
Apps & games
Art

Open weightsClosed

Rank

ModelOpen weightsClosed

95% confidence interval80010001200

Open

ProvisionalRanked once it has 25 votes and a tight range

Provisional, not ranked yet

Fable 5.1Anthropic, closed

1177

1000 to 1268

100%

2

$4.73

Provisional, not ranked yet

Kimi K3Moonshot AI, open weights

1012

869 to 1157

50%

2

$0.21

Provisional, not ranked yet

GLM-5.3Z.ai, open weights

929

852 to 1002

0%

1

$0.30

Provisional, not ranked yet

DeepSeek V4 FlashDeepSeek, open weights

882

780 to 1000

0%

1

$0.01

ProvisionalRanked once it has 25 votes and a tight range
Provisional, not ranked yetFable 5.1Anthropic, closed1177100% won95% interval 1000 to 1268, 2 votes
Provisional, not ranked yetKimi K3Moonshot AI, open weights101250% won95% interval 869 to 1157, 2 votes
Provisional, not ranked yetGLM-5.3Z.ai, open weights9290% won95% interval 852 to 1002, 1 votes
Provisional, not ranked yetDeepSeek V4 FlashDeepSeek, open weights8820% won95% interval 780 to 1000, 1 votes

Showing 4 models, Stories, table view.

Bradley-Terry scores, 95% intervals. Overlapping intervals share a rank.
Side A won 46.9% of 32 decisive votes.
Ranked at 100 votes (25 per category).
JSON