Arena WebDev · React

Human preference between agentic front-end web development outputs; Bradley-Terry rating. Only the React category.

All results

110 rows

#	Model	Effort	Arena WebDev · React
1	Opus 5.5	max	1860 ±16
2	Sonnet 5.5	xhigh	1817 ±14
3	Fable 5.1	max	1792 ±11
4	Gemini 4 Carbon	high	1791 ±18
5	GPT-6 Astra	max	1790 ±12
6	GPT-6.1 Sol	max	1752 ±14
7	Sonnet 5.5	high	1748 ±11
8	Opus 5	max	1714 ±7
9	Kimi K3.1	max	1701 ±16
10	Gemini 4 Argon	high	1694 ±14
11	Qwen 3.8 Max	max	1692 ±8
12	GPT-6 Sol	max	1680 ±12
13	Opus 5	high	1678 ±6
14	DeepSeek V4.1 Flash	max	1673 ±10
15	Kimi K3	max	1672 ±7
16	Muse Spark 1.3	max	1670 ±9
17	Grok 4.7	xhigh	1663 ±11
18	GPT-6.1 Terra	max	1657 ±19
19	Qwen 3.8 Flash	max	1657 ±9
20	Hy4	high	1653 ±12
21	MiMo V2.6 Flash	high	1651 ±10
22	GLM 5.3	max	1643 ±9
23	MiMo V2.6 Pro	high	1635 ±10
24	Muse Spark 1.3	xhigh	1631 ±9
25	Fable 5	high	1630 ±7
26	Grok 4.6	high	1630 ±8
27	GLM 5.3 Flash	max	1624 ±7
28	GLM 5.2	max	1619 ±7
29	GPT-5.6 Sol	max	1618 ±7
30	Haiku 5.5	high	1608 ±12