BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases9

Arena WebDev

Pairwise votes on generated web apps.

As of 26 Oct 2026, Claude Opus 5.5 leads Arena WebDev on BenchLeader with 1807, ahead of GPT-6 Astra at 1774, across 139 model configurations with a published result.

Published by
Arena (LMArena)
Category
Coding
Index weight
1.0
Models
139
Data as of
26 Oct 2026
Newest result
23 Oct 2026

CC BY 4.0 — lmarena-ai/leaderboard-dataset on Hugging Face.

What the test looks like

LMArena WebDev asks two models to build the same small web app from a prompt, renders both, and lets the visitor vote on the working result.

How it is scored

Bradley-Terry rating from pairwise votes on the rendered apps.

What to keep in mind

Votes reward visual polish and speed of a first draft; a broken build loses regardless of code quality. Fewer votes than the text arena.

Top 15
  1. 1Claude Opus 5.5 (max)1807
  2. 2GPT-6 Astra (max)1774
  3. 3Claude Sonnet 5.5 (xhigh)1769
  4. 4GPT-6.1 Sol (max)1749
  5. 5Claude Fable 5.1 (max)1743
  6. 6Muse Wonder Full (max)1726
  7. 7Claude Sonnet 5.5 (high)1709
  8. 8Kimi K3.1 Pro (max)1692
  9. 9Claude Opus 5 (max)1691
  10. 10GPT-6 Sol (max)1681
  11. 11Gemini 4 Argon (high)1674
  12. 12Qwen3.8 Max (max)1673
  13. 13DeepSeek V4.1 Flash 1017 (max)1665
  14. 14Claude Opus 5 (high)1659
  15. 15Muse Spark 1.3 (max)1657
139 of 139
#
1Claude Opus 5.5max180772.62026-10-23
2GPT-6 Astramax177470.42026-10-23
3Claude Sonnet 5.5xhigh176967.92026-10-23
4GPT-6.1 Solmax174969.92026-10-23
5Claude Fable 5.1max174370.52026-10-23
6Muse Wonder Fullmax172670.62026-10-23
7Claude Sonnet 5.5high170965.62026-10-23
8NewKimi K3.1 Promax1692–2026-10-23
9Claude Opus 5max169168.82026-10-23
10GPT-6 Solmax168166.42026-10-23
11Gemini 4 Argonhigh167470.72026-10-23
12Qwen3.8 Maxmax167363.32026-10-23
13DeepSeek V4.1 Flash 1017max166564.02026-10-23
14Claude Opus 5high165967.92026-10-23
15Muse Spark 1.3max165767.12026-10-23
16Kimi K3max165564.82026-10-23
17Muse Wondermax165167.52026-10-23
18Grok 4.7xhigh163861.62026-10-23
19MiMo-V2.6-Flash163559.72026-10-23
20Qwen3.8-Flash-Next162960.32026-10-23
21MiMo-V2.6-Pro162564.02026-10-23
22Claude Fable 5high162464.42026-10-23
23GLM 5.3max162363.52026-10-23
24Muse Spark 1.3xhigh162064.92026-10-23
25DeepSeek V4.1 Flashmax161961.22026-10-23
26Grok 4.6high161762.92026-10-23
27GPT-5.6 Solxhigh161566.62026-10-23
28GLM 5.3 Flash160862.72026-10-23
29GLM 5.2max160462.32026-10-23
30Qwen3.8 27B159154.62026-10-23
31Gemini 3.7 Flashhigh159063.52026-10-23
32NewMiniMax M3.1 Flash158961.92026-10-23
33Claude Haiku 5.5high158659.62026-10-23
34Gemini 3.8 Flashhigh158564.42026-10-23
35DeepSeek V4 Prohigh158257.52026-10-23
36DeepSeek V4 Flashhigh158257.32026-10-23
37GPT-6 Lunamax157960.02026-10-23
38Step 5 Previewhigh1568–2026-10-23
39Claude Opus 4.7high155861.02026-10-23
40Claude Opus 4.8high155660.72026-10-23
41Claude Opus 4.7155563.22026-10-23
42Grok 4.5155359.02026-10-23
43Claude Opus 4.6high154662.72026-10-23
44Mistral Large 4154556.72026-10-23
45Muse Spark 1.1154362.52026-10-23
46Claude Sonnet 5high154160.12026-10-23
47Gemini 3.6 Flashhigh153862.02026-10-23
48Claude Opus 4.6153762.52026-10-23
49Claude Opus 4.8153660.12026-10-23
50Muse Spark 1.2xhigh153363.22026-10-23
51Muse Spark Openmax1526–2026-10-23
52Claude Sonnet 4.6152357.12026-10-23
53GPT-5.6 Terraxhigh152162.82026-10-23
54Seed2.1 Pro1520–2026-10-23
55GPT-5.6 Lunaxhigh151959.72026-10-23
56Qwen3.7max151560.62026-10-23
57GPT-5.5xhigh151366.12026-10-23
58Kimi K2.6150955.32026-10-23
59GLM 5.1150855.92026-10-23
60Gemini 3.5 Flashhigh149963.12026-10-23

More coding benchmarks · All LLM benchmarks

Cite as: BenchLeader, “Arena WebDev leaderboard”, data as of 26 Oct 2026, https://www.benchleader.com/benchmarks/lmarena_webdev.

Arena WebDev: questions

What does Arena WebDev measure?
LMArena WebDev asks two models to build the same small web app from a prompt, renders both, and lets the visitor vote on the working result. Scores are reported in Elo-style ratings from pairwise comparisons; higher is better.
Which AI model leads Arena WebDev?
Claude Opus 5.5 leads Arena WebDev with 1807 as of 26 Oct 2026, ahead of GPT-6 Astra at 1774.
How many models have Arena WebDev results?
139 model configurations have a Arena WebDev result on BenchLeader, all taken from Arena (LMArena).
Who runs Arena WebDev and how often is it updated?
Arena WebDev is published by Arena (LMArena). BenchLeader re-reads the published results once a day and records the date each result was published.
Does Arena WebDev count toward the BenchLeader Index?
Yes. Arena WebDev contributes to the coding category of the BenchLeader Index, normalised so that 50 is the average of the evaluated models.