DesignArena: Full-Stack

Two coding agents each build a full-stack web app from the same real user prompt, a React app with sign-in, a database, and storage on Supabase, inside DesignArena’s sandboxed harness. A person compares the two finished apps and votes for one, and a Bradley–Terry fit over every vote rates each model on an Elo scale that starts at 1200. Cost is each model’s list price per million tokens, not spend per build.

Last updated · Source leaderboard checked daily

Main Insights

Results

DesignArena: Full-Stack results, ranked by score (elo), highest first. Last updated .
RankModelProviderScore (Elo)Cost (USD / task)
1Claude Fable 5.2Anthropic1402.0%$16.00
2Claude Opus 5.5Anthropic1389.0%$8.00
3GPT-6 Astra (xhigh)OpenAI1331.0%$20.00
4Kimi K3.1 ProMoonshot AI1318.0%$4.00
5Claude Fable 5.1Anthropic1299.0%$20.00
6Kimi K3Moonshot AI1288.0%$6.00
7Muse Spark 1.3 MaxMeta1286.0%$2.00
8Claude Opus 5Anthropic1280.0%$10.00
9Qwen 3.8 MaxAlibaba1277.0%$3.00
10Claude Fable 5Anthropic1253.0%$20.00
11Grok 4.6xAI1236.0%$3.00
12GPT-5.6 Sol (xhigh)OpenAI1231.0%$8.00
13Claude Sonnet 5Anthropic1222.0%$4.00
14Claude Opus 4.8Anthropic1222.0%$10.00
15Grok 4.5xAI1221.0%$3.00