MODEL INTELLIGENCE

ModelRank

Compare models within one task and one source. Human preference, benchmark ability, and usage measure different things; their scores are not combined.

Source LMArena WebDev Source publication: Oct 28, 2026 · Days since publication: 6 Last successful source check: 2026-11-03 13:00 UTC Snapshot stored: 2026-11-03 04:05 UTC

Original source

Methodology and data notes

Blind human preference for web app outputs. This measures user preference, not unit-test pass rate or general programming correctness.

The source contains conflicting results for the same model label. These ambiguous entries are excluded; other source ranks are preserved.

Metric: Arena rating, higher is better

Other coding benchmark: official SWE-bench leaderboard

Ranking source

LMArena WebDev Arena rating LMArena Agent IPS score

Rank	Model	Organization	Arena rating ↑	95% CI	Access	Evidence
1	Claude Opus 5.5 (Max) Arena rating ↑	anthropic	1,809.8	1797.0–1822.7	Proprietary	3,473 votes Source
2	GPT-6 Astra (Max) Arena rating ↑	openai	1,773.4	1763.8–1782.9	Proprietary	7,210 votes Source
3	Claude Sonnet 5.5 (xHigh) Arena rating ↑	anthropic	1,769.7	1758.5–1781.0	Proprietary	4,469 votes Source
4	Gemini 4 Carbon (High) Arena rating ↑	google	1,768.2	1753.2–1783.2	Proprietary	2,230 votes Source
5	GPT-6.1 Sol (Max) Arena rating ↑	openai	1,750.6	1739.1–1762.0	Proprietary	4,092 votes Source
6	Claude Fable 5.1 (Max) Arena rating ↑	anthropic	1,745.9	1737.3–1754.4	Proprietary	8,614 votes Source
7	Muse Wonder Full (Max) Arena rating ↑	meta	1,725.3	1711.5–1739.2	Proprietary	2,248 votes Source
8	Claude Sonnet 5.5 (High) Arena rating ↑	anthropic	1,710.6	1701.4–1719.8	Proprietary	6,371 votes Source
9	Kimi K3.1 Pro (Max) Arena rating ↑	moonshot	1,694.6	1681.6–1707.5	Proprietary	2,671 votes Source
10	Claude Opus 5 (Max) Arena rating ↑	anthropic	1,690.5	1684.4–1696.6	Proprietary	19,641 votes Source
11	Gemini 4 Argon (High) Arena rating ↑	google	1,679.2	1667.8–1690.6	Proprietary	3,437 votes Source
12	GPT-6 Sol (Max) Arena rating ↑	openai	1,678.2	1668.8–1687.6	Proprietary	5,714 votes Source
13	Qwen3.8-Max (0902) Arena rating ↑	alibaba	1,675.9	1669.3–1682.6	Proprietary	17,936 votes Source
14	GPT-6.1 Terra (Max) Arena rating ↑	openai	1,672.7	1657.3–1688.0	Proprietary	1,821 votes Source
15	Qwen3.8-Max Arena rating ↑	alibaba	1,671.9	1659.8–1684.0	Proprietary	3,406 votes Source
16	DeepSeek-V4.1-Flash-20261017 (Max) Arena rating ↑	deepseek	1,665.0	1657.0–1673.1	MIT	7,842 votes Source
17	Claude Opus 5 (High) Arena rating ↑	anthropic	1,658.2	1652.9–1663.5	Proprietary	26,887 votes Source
18	Kimi K3 (Max) Arena rating ↑	moonshot	1,656.4	1650.6–1662.2	Kimi K3 license	21,281 votes Source
19	Muse Spark 1.3 (Max) Arena rating ↑	meta	1,656.2	1648.8–1663.7	Proprietary	10,796 votes Source
20	Muse Wonder (Max) Arena rating ↑	meta	1,652.1	1641.0–1663.2	Proprietary	3,724 votes Source
21	Grok 4.7 (xHigh) Arena rating ↑	xai	1,637.5	1628.7–1646.2	Proprietary	6,572 votes Source
22	MiMo-V2.6-Flash Arena rating ↑	xiaomi	1,637.2	1628.7–1645.6	MIT	7,068 votes Source
23	Hy4 preview Arena rating ↑	tencent	1,631.1	1621.5–1640.8	Apache 2.0	4,765 votes Source
24	Qwen3.8-Flash-Next Arena rating ↑	alibaba	1,628.3	1620.7–1635.9	qwen-community-1.0	9,703 votes Source
25	MiMo-V2.6-Pro Arena rating ↑	xiaomi	1,625.4	1617.0–1633.8	MIT	6,820 votes Source
26	Claude Fable 5 (High) Arena rating ↑	anthropic	1,624.1	1618.1–1630.1	Proprietary	16,951 votes Source
27	GLM-5.3 (Max) Arena rating ↑	zai	1,622.8	1615.6–1630.1	MIT	11,150 votes Source
28	Muse Spark 1.3 (xHigh) Arena rating ↑	meta	1,620.1	1612.7–1627.5	Proprietary	10,243 votes Source
29	DeepSeek-V4.1-Flash-20260910 (Max) Arena rating ↑	deepseek	1,619.2	1608.2–1630.1	MIT	3,510 votes Source
30	Grok 4.6 (High) Arena rating ↑	xai	1,616.4	1609.4–1623.4	Proprietary	11,966 votes Source
31	GPT-5.6 Sol (xHigh, Codex Harness) Arena rating ↑	openai	1,615.8	1610.0–1621.5	Proprietary	18,884 votes Source
32	GLM-5.3-Flash Arena rating ↑	zai	1,606.6	1600.3–1613.0	MIT	21,587 votes Source
33	GLM-5.2 (Max) Arena rating ↑	zai	1,605.6	1599.7–1611.5	MIT	17,817 votes Source
34	Qwen3.8-27B Arena rating ↑	alibaba	1,589.8	1583.5–1596.0	Apache 2.0	18,752 votes Source
35	Gemini 3.7 Flash (High) Arena rating ↑	google	1,589.7	1583.2–1596.2	Proprietary	14,473 votes Source
36	Claude Haiku 5.5 (High) Arena rating ↑	anthropic	1,588.8	1579.2–1598.5	Proprietary	4,758 votes Source
37	Gemini 3.8 Flash (High) Arena rating ↑	google	1,583.3	1576.3–1590.3	Proprietary	12,342 votes Source
38	MiniMax-M3.1-Flash Arena rating ↑	minimax	1,582.5	1570.9–1594.0	MiniMax Community License	3,062 votes Source
39	DeepSeek-V4-Pro (High, 20260813) Arena rating ↑	deepseek	1,581.9	1572.4–1591.3	MIT	4,676 votes Source
40	DeepSeek-V4-Flash (High) Arena rating ↑	deepseek	1,581.8	1572.5–1591.1	MIT	4,927 votes Source
41	GPT-6 Luna (Max) Arena rating ↑	openai	1,579.4	1573.6–1585.2	Proprietary	27,808 votes Source
42	Step 5 Preview (High) Arena rating ↑	stepfun	1,571.0	1562.1–1580.0	Proprietary	5,966 votes Source
43	Claude Opus 4.7 (High) Arena rating ↑	anthropic	1,557.5	1551.8–1563.3	Proprietary	17,859 votes Source
44	Claude Opus 4.8 (High) Arena rating ↑	anthropic	1,555.8	1550.1–1561.6	Proprietary	18,408 votes Source
45	Claude Opus 4.7 Arena rating ↑	anthropic	1,555.4	1549.8–1561.1	Proprietary	18,167 votes Source
46	Grok 4.5 Arena rating ↑	xai	1,553.0	1546.4–1559.7	Proprietary	11,018 votes Source
47	Claude Opus 4.6 (High) Arena rating ↑	anthropic	1,546.2	1540.8–1551.7	Proprietary	19,316 votes Source
48	Mistral Large 4 Arena rating ↑	mistral	1,544.3	1533.8–1554.9	Proprietary	4,068 votes Source
49	Muse Spark 1.1 Arena rating ↑	meta	1,542.5	1534.9–1550.1	Proprietary	7,935 votes Source
50	Claude Sonnet 5 (High) Arena rating ↑	anthropic	1,540.1	1534.3–1545.9	Proprietary	16,832 votes Source