Atlas

Benchmarks

← All benchmarks

Design Arena Overall Code Categories

Code

Design Arena Overall Code Categories aggregates blind, pairwise human preferences across Website, UI Component, Game Development, Data Visualization, and 3D tasks, each produced as a single-file HTML output. The leaderboard converts head-to-head results into Bradley–Terry Elo ratings.

Top models (higher is better)

ModelScore
Claude Opus 5.51385
GPT-6 Astra1384
Muse Wonder1381
Kimi K31359
GPT-6.1 Sol1357
Kimi K3.1 Pro1354
Claude Sonnet 5.51352
Muse Spark 1.31351
GPT-5.6 Sol1340
Claude Fable 5.11334
Gemini 4 Argon1330
Claude Opus 51323
DeepSeek V4.1 Flash1318
MiMo-V2.6-Pro1316
Muse Spark 1.21311
Claude Fable 51309
Gemini 3.7 Flash1306
Ling 3.1 Flash1303
Gemini 3.8 Flash1302
Muse Spark Open1300
GLM-5.31299
MiniMax M3.1 Flash1298
GPT-6.1 Terra1296
GPT-6 Sol1296
Grok 4.61290
DeepSeek V4.1 Flash 10171289
Gemini 3.6 Flash1289
Opus 4.61287
GLM-5.21285
Claude Haiku 5.51280
GLM-5.3 Flash1279
Muse Spark 1.11275
Sonnet 4.61274
Sonnet 51274
Grok 4.51274
Qwen3.7-Max (source unspecified)1274
Grok 4.71270
Kimi K2.61270
GLM-5.11267
Qwen3.7-Plus (source unspecified)1267
Loading Atlas data…