Atlas

Benchmarks

← All benchmarks

Arena Code WebDev — Fullstack

Code · 2026-07-28

Arena Code WebDev rating for full-stack web development tasks: multi-step reasoning, tool use and end-to-end app generation. Anonymous generated applications are compared through blind pairwise user votes; with Frontend it splits the overall WebDev votes.

Top models (higher is better)

ModelScore
GPT-6 Astra1753
Claude Opus 5.51739
GPT-6.1 Sol1736
Muse Wonder Full1717
GPT-6 Sol1703
Claude Sonnet 5.51695
Qwen3.8-Max1689
Claude Fable 5.11688
DeepSeek V4.1 Flash 10171676
Muse Wonder1674
Qwen3.8 Max (0902)1673
Gemini 4 Argon1672
Claude Opus 51668
Kimi K31657
Muse Spark 1.31653
MiMo-V2.6-Pro1631
GPT-5.6 Sol1630
Gemini 3.7 Flash1629
GLM-5.31625
DeepSeek V4.1 Flash1623
Grok 4.71619
MiMo-V2.6-Flash1619
Grok 4.61616
Claude Fable 51616
Qwen3.8 27B1615
DeepSeek-V4-Flash1614
Gemini 3.8 Flash1614
GLM-5.21613
Qwen3.8-Flash-Next1611
GLM-5.3 Flash1606
Hy4 Preview1604
GPT-6 Luna1602
MiniMax M3.1 Flash1600
DeepSeek V4 Pro 08131598
Muse Spark 1.11594
Gemini 3.6 Flash1582
Opus 4.81580
Opus 4.61576
Sonnet 4.61576
Claude Haiku 5.51570
Loading Atlas data…