The honest view from inside technical leadership
LLM Leaderboard · Coding (human-judged)

WebDev Arena leaderboard

Head-to-head votes from developers who ask two anonymous models to build the same web app and pick the better result. Votes are turned into an Elo-style rating.

← All models

Rankings

Snapshot 22 November 2026 · Scores copied from each benchmark's own leaderboard · How the CTO Index works

Score and price

#ModelProviderWebDevPrice $/M
1Claude Fable 5.2Anthropic1824$8 / $40
2Claude Opus 5.5Anthropic1819$4 / $20
3Claude Sonnet 5.5Anthropic1772$2 / $10
4Gemini 4 CarbonGoogle1769$2 / $10
5GPT-6.2 AstraOpenAI1766$8 / $40
6Gemini 4 NeonGoogle1765$1.50 / $7.50
7GPT-6 AstraOpenAI1764$10 / $50
8GPT-6.2 SolOpenAI1754$2 / $10
9Qwen4-MaxAlibaba1751$1.50 / $4.50
10GPT-6.1 SolOpenAI1743$2 / $10
11Claude Fable 5.1Anthropic1742$10 / $50
12Muse Wonder FullMeta1723Not listed
13Gemini 4 HydrogenGoogle1701$1.50 / $7.50
14Claude Opus 5Anthropic1692$5 / $25
15GPT-6.1 TerraOpenAI1685$1 / $5
16Kimi K3.1 ProMoonshot AI1683$2 / $10
17Gemini 4 ArgonGoogle1682$2 / $10
18GPT-6 SolOpenAI1674$2 / $10
19Qwen3.8-MaxAlibaba1674$2 / $6
20Kimi K3Moonshot AI1659$3 / $15
21DeepSeek V4.1 FlashDeepSeek1658$0.15 / $0.60
22Muse Spark 1.3Meta1653$1.25 / $4.25
23Hy4Tencent1650Not listed
24Muse WonderMeta1647$4 / $20
25Inkling 2 SmallThinking Machines1643$0.30 / $1
26Grok 4.7xAI1639$2 / $6
27MiMo-V2.6-FlashXiaomi1638$0.14 / $0.28
28Claude Fable 5Anthropic1626$10 / $50
29GLM-5.3Z.ai1623$1.40 / $4.40
30MiMo-V2.6-ProXiaomi1622$0.43 / $0.87
31Grok 4.6xAI1618$1.60 / $4.80
32GPT-5.6 SolOpenAI1615$4 / $20
33GLM-5.3-FlashZ.ai1610$0.02 / $0.07
34Muse EmberMeta1608$0.10 / $0.50
35Claude Haiku 5.5Anthropic1595$0.10 / $0.50
36Qwen3.8-27BAlibaba1590$0.40 / $3
37Gemini 3.7 FlashGoogle1589$0.75 / $3.75
38Gemini 3.8 FlashGoogle1584$0.75 / $3.75
39GPT-6 LunaOpenAI1578$0.10 / $0.50
40Mistral Large 4Mistral1573$0.70 / $2.10
41Step 5 PreviewStepFun1570$1 / $2.70
42Claude Opus 4.8Anthropic1555$5 / $25
43Claude Sonnet 5Anthropic1539$2 / $10
44Beam 1.0Reflection AI1437$0.30 / $1

Other benchmarks

Frequently asked questions

What does WebDev Arena measure?

Head-to-head votes from developers who ask two anonymous models to build the same web app and pick the better result. Votes are turned into an Elo-style rating.

Which model has the highest WebDev Arena score?

Claude Fable 5.2 from Anthropic leads WebDev Arena at 1824, ahead of Claude Opus 5.5 (1819).

Why does WebDev Arena matter to engineering leaders?

It captures what engineers actually prefer when they look at working output, including UI quality and following the brief, which pass/fail tests miss.

What are the limitations of WebDev Arena?

Preference votes favour visually polished front-end work and say little about back-end or infrastructure code. New models carry wider confidence intervals until they collect votes.