llm.ing

Design Arena — fullstack

mirrorrun by Design Arenaunit: elo32 modelssource ↗

Caveat: Head-to-head human votes on generated designs/code, mirrored via OpenRouter.

websitecodecategoriesdatavizuicomponentgamedev3dsvgasciiartwebappsfullstack+15 more
#ModelScoreObserved
1Claude Opus 4.7
1503
Jul 23, 2026
2Claude Fable 5
1319
Jul 23, 2026
3Claude Opus 4.8
1307
Jul 23, 2026
4Grok 4.5
1301
Jul 24, 2026
5Claude Sonnet 5
1290
Jul 24, 2026
6GLM-5.2
1288
Jul 23, 2026
7Claude Opus 4.6
1278
Jul 23, 2026
8Claude Sonnet 4.6
1267
Jul 23, 2026
9Gemini 3.5 Flash
1252
Jul 24, 2026
10MiniMax-M3
1250
Jul 23, 2026
11Kimi K2.7 Code
1239
Jul 24, 2026
12Qwen3.7 Max
1225
Jul 23, 2026
13GLM-5.1
1224
Jul 23, 2026
14Claude Opus 4.5 (latest)
1215
Jul 23, 2026
15Kimi K2.6
1214
Jul 24, 2026
16GLM-5V-Turbo
1201
Jul 23, 2026
17GLM-5
1179
Jul 23, 2026
18Kimi K2.5
1172
Jul 23, 2026
19GPT-5.5
1142
Jul 23, 2026
20Gemini 3.1 Pro Preview
1126
Jul 23, 2026
21Gemini 3 Flash Preview
1120
Jul 23, 2026
22Claude Sonnet 4.5 (latest)
1111
Jul 23, 2026
23Grok 4.20 (Non-Reasoning)
1110
Jul 23, 2026
24GLM-4.7
1107
Jul 23, 2026
25GPT-5.2
1100
Jul 23, 2026
26GLM-4.6
1088
Jul 23, 2026
27GPT-5.1 Codex
1086
Jul 23, 2026
28GPT-5.4
1075
Jul 23, 2026
29Grok 4.3
1069
Jul 24, 2026
30GPT-5.2 Codex
1050
Jul 23, 2026
31GPT-5.3 Codex
1040
Jul 23, 2026
32DeepSeek V4 Pro
948
Jul 23, 2026

Dot = elo on a 920–1531 scale (position, not length). Scores link to model pages; every number keeps its source on the model page.