WebDev Arena Leaderboard

WebDev Arena is an open-source benchmark evaluating AI capabilities in web development, developed by LMArena.

Leaderboard

Claude 3.5 Sonnet (20241022)

Anthropic

Arena Score

1252.24

License

Proprietary

95% CI

+4.91 / -6.09

Votes

12,356

Claude 3.5 Haiku (20241022)

Anthropic

Arena Score

1138.14

License

Proprietary

95% CI

+6.33 / -6.88

Votes

6,288

o1-mini (20240912)

OpenAI

Arena Score

1068.48

License

Proprietary

95% CI

+6.10 / -5.58

Votes

9,224

o1 (20241217)

OpenAI

Arena Score

1060.55

License

Proprietary

95% CI

+7.32 / -9.25

Votes

4,463

Gemini-2.0-Flash-Thinking-1219

Google

Arena Score

1029.45

License

Proprietary

95% CI

+6.21 / -5.47

Votes

8,010

Gemini-Exp-1206

Google

Arena Score

1025.87

License

Proprietary

95% CI

+5.42 / -4.43

Votes

11,388

Gemini-2.0-Flash-Exp

Google

Arena Score

978.07

License

Proprietary

95% CI

+4.95 / -4.99

Votes

11,046

DeepSeek v3

DeepSeek

Arena Score

967.88

License

DeepSeek

95% CI

+8.46 / -9.30

Votes

5,345

GPT-4o-2024-11-20

OpenAI

Arena Score

964.00

License

Proprietary

95% CI

+4.16 / -4.89

Votes

11,977

Qwen2.5-Coder-32B-Instruct

Alibaba

#10

Arena Score

902.78

License

Apache 2.0

95% CI

+6.16 / -5.16

Votes

10,788

Gemini-1.5-Pro-002

Google

#10

Arena Score

894.46

License

Proprietary

95% CI

+5.42 / -4.77

Votes

10,488

Llama-3.1-405B-Instruct

Meta

#12

Arena Score

811.76

License

Llama 3.1

95% CI

+13.01 / -16.72

Votes

1,117

Rank (UB)	Model	Arena Score	95% CI	Votes	Organization	License
1	Claude 3.5 Sonnet (20241022)	1252.24	+4.91 / -6.09	12,356	Anthropic	Proprietary
2	Claude 3.5 Haiku (20241022)	1138.14	+6.33 / -6.88	6,288	Anthropic	Proprietary
3	o1-mini (20240912)	1068.48	+6.10 / -5.58	9,224	OpenAI	Proprietary
3	o1 (20241217)	1060.55	+7.32 / -9.25	4,463	OpenAI	Proprietary
5	Gemini-2.0-Flash-Thinking-1219	1029.45	+6.21 / -5.47	8,010	Google	Proprietary
5	Gemini-Exp-1206	1025.87	+5.42 / -4.43	11,388	Google	Proprietary
7	Gemini-2.0-Flash-Exp	978.07	+4.95 / -4.99	11,046	Google	Proprietary
7	DeepSeek v3	967.88	+8.46 / -9.30	5,345	DeepSeek	DeepSeek
8	GPT-4o-2024-11-20	964.00	+4.16 / -4.89	11,977	OpenAI	Proprietary
10	Qwen2.5-Coder-32B-Instruct	902.78	+6.16 / -5.16	10,788	Alibaba	Apache 2.0
10	Gemini-1.5-Pro-002	894.46	+5.42 / -4.77	10,488	Google	Proprietary
12	Llama-3.1-405B-Instruct	811.76	+13.01 / -16.72	1,117	Meta	Llama 3.1

More Statistics for WebDev Arena (Overall)

Confidence Interval for Model Strength

Figure 1

Average Win Rate Against All Other Models (Assuming Uniform Sampling and No Ties)

Figure 2

Fraction of Model A Wins for All Non-tied A vs. B Battles

Figure 3

Battle Count for Each Combination of Models (without Ties)

Figure 4