Atlas

Benchmarks

← All benchmarks

Arena Code WebDev — Frontend

Code

Arena Code WebDev rating for front-end web development tasks: multi-step reasoning, tool use and HTML or React code generation. It holds most overall WebDev votes and nearly duplicates the overall board, so it is a component of the WebDev family.

Top models (higher is better)

ModelScore
Claude Opus 5.51849
Claude Sonnet 5.51808
GPT-6 Astra1795
Claude Fable 5.11776
GPT-6.1 Sol1767
Muse Wonder Full1742
Claude Opus 51712
Kimi K3.1 Pro1704
Gemini 4 Argon1688
GPT-6 Sol1687
Qwen3.8 Max (0902)1686
Qwen3.8-Max1679
DeepSeek V4.1 Flash 10171675
Muse Spark 1.31672
Kimi K31669
Grok 4.71660
Muse Wonder1657
MiMo-V2.6-Flash1652
Hy4 Preview1651
Qwen3.8-Flash-Next1648
GLM-5.31639
Claude Fable 51637
MiMo-V2.6-Pro1637
DeepSeek V4.1 Flash1630
Grok 4.61629
GPT-5.6 Sol1623
GLM-5.3 Flash1621
GLM-5.21614
Claude Haiku 5.51604
MiniMax M3.1 Flash1601
Qwen3.8 27B1598
Gemini 3.7 Flash1592
DeepSeek V4 Pro 08131590
Gemini 3.8 Flash1588
GPT-6 Luna1586
DeepSeek-V4-Flash1585
Opus 4.71562
Grok 4.51562
Opus 4.81560
Mistral Large 4 Preview1553
Loading Atlas data…