Run 001 · 10 models · 7 challenges

See what models can actually build.

Identical prompts. One generation each. No human edits. Browse the live HTML by use case or by model, then upvote the strongest artifact.

Open visual arena Browse model library

Frontier models: 10
Visual challenges: 7
Published builds: 70
Human edits: 0

Visual arena

Compare one-shot builds by use case

Challenge

Solar System · Geometry / depth

Open challenge page

Grok 4.6

xAI

0— upvoted

GPT-5.6 Luna

OpenAI

0— upvoted

Gemini 3.8 Flash

Google

0— upvoted

GLM 5.3 Flash

Z.ai

0— upvoted

HY 4 Preview

Tencent

0— upvoted

DeepSeek V4 Flash 0731

DeepSeek

0— upvoted

Kimi K3

Moonshot AI

0— upvoted

GPT-5.6 Sol

OpenAI

0— upvoted

Muse Spark 1.3

Meta

0— upvoted

Gemma 4 31B IT

Google

0— upvoted

Methodology

Transparent, one-shot, comparable.

Identical prompt

Every model receives the same task and system constraints.

Published as generated

No cherry-picked variants. Infra retries are labeled in the manifest.

Native web only

One HTML file. HTML, CSS, SVG, Canvas, and vanilla JavaScript.

Community ranking

Reddit-style upvotes compute score and percent upvoted per artifact.