Atlas

Benchmarks

← All benchmarks

LMArena Code Arena WebDev

Code · 2024-12-10

LMArena Code Arena WebDev is an arena-style leaderboard where users compare generated websites and web apps from anonymous model outputs. Scores are reported as Elo ratings for the web-development track.

Top models (higher is better)

ModelScore
DeepSeek-V4-Flash1577
Opus 4.81566
Opus 4.71562
Opus 4.61555
Grok 4.51551
Qwen3.7-Max1541
Hy31517
Muse Spark1513
Opus 4.51512
GPT-5.51505
Gemini 3.5 Flash1492
Qwen3.6-Max-Preview1486
GPT-5.21480
GPT-5.41457
DeepSeek-V4-Pro1447
Kimi K2.51436
Qwen3.5 397B A17B1400
GLM 5V Turbo1399
Grok 4.201395
Qwen3.5 Plus (2026-02-15)1393
MiniMax M2.11392
Sonnet 4.51391
Opus 4.11386
Gemini 3 Flash Preview1383
Gemma 4 31B IT1380
DeepSeek-V3.2-Exp1367
Gemma 4 26B A4B1367
Gemma 4 31B1363
DeepSeek-V3.21361
Qwen3.5 122B A10B1360
GPT-5.11359
Qwen3.5-27B1357
MiMo-V2-Flash1337
Haiku 4.51286
Mistral Medium 3.51267
Gemini 3.1 Flash-Lite Preview1256
Qwen3.6 35B A3B1249
Gemini 3.1 Flash-Lite1248
GPT-5.1-Codex mini1247
Mistral Large 3 675B Instruct 25121230
Loading Atlas data…