benchmarks.site

Public leaderboard · Human-preference arenas

Design Arena

People ask several AI models to build the same thing, such as a website, game or slide deck, then vote on which result looks and works better. The votes become a ranking of which models are best at making designs and apps.

Arena (people vote)

Top 3 right now

  1. 1stClaude Opus 5.5 (high)1445
  2. 2ndGPT-6 Astra (max)1399
  3. 3rdClaude Fable 5.21367

What it measures

Crowdsourced votes comparing AI-generated designs, including websites, games, slides, images, video and audio, with separate leaderboards per output type.

How to read the numbers

Elo rating from head-to-head user votes (higher means the model's output was preferred more often) — higher is better.

Results

15 rows · higher is better (Elo)CSV
1Claude Opus 5.5 (high)Anthropic144569.1%2,670
2GPT-6 Astra (max)OpenAI139967.2%2,526
3Claude Fable 5.2Anthropic136769.0%1,738
4GPT-6.1 Sol (max)OpenAI136651.3%2,069
5Qwen4-MaxAlibaba136062.5%3,474
6Claude Fable 5.1Anthropic133659.5%5,155
7Inkling 2 SmallThinking Machines131157.2%7,217
8Gemini 4 CarbonGoogle130660.5%1,054
9Kimi K3.1 ProMoonshot AI129759.4%4,489
10Claude Sonnet 5.5Anthropic129057.5%3,030
11DeepSeek-V4.1-Flash (1017)DeepSeek128847.8%404
12Muse Spark 1.3 MaxMeta127956.9%13,091
13Grok 4.7xAI127354.4%1,506
14Claude Fable 5Anthropic127260.5%13,563
15Gemini 4 HydrogenGoogle126856.6%380

Note: 'Overall Agentic' board (combines web apps, full-stack, game dev, React Native, Android, slides and spreadsheets). Only the first 15 rows are shown on the page without expanding; separate boards exist per output type. Page says 'Updated Nov 17, 2026, 10:31 AM UTC'.

Source: DesignArena · designarena.ai · results as of 2026-11-17 · retrieved 2026-11-17

Markdown · More Human-preference arenas