M
ModelSage
排行榜/Arena 人类盲测/Code 编程/品牌营销站

Arena Code 编程 · 品牌营销站

前端 / 全栈代码生成盲测，含 11 个细分场景子榜 · 完整榜单 142 个模型

Overall 总榜143Fullstack 全栈82Frontend 前端143HTML142React127品牌营销站142参考图还原142数据分析应用142消费级产品142游戏142仿真模拟142内容创作工具138
RankModelElo 分
1
claude-opus-5.5-max1803.4±25.4
2
claude-sonnet-5.5-xhigh1764.1±24.6
3
gpt-6-astra-max1737.8±19
4
claude-fable-5.1-max1737.6±15.6
5
gpt-6.1-sol-max1710.2±24.8
6
claude-opus-5-max1680.5±10.4
7
kimi-k3-max1679.2±9.9
8
qwen3.8-max1670.5±21.3
9
claude-sonnet-5.5-high1669±18.9
10
qwen3.8-max-09021667.4±11.8
11
gpt-6-sol-max1664.1±20.7
12
muse-wonder-max1656.4±35.5
13
muse-spark-1.3-max1646±13.8
14
grok-4.7-xhigh1645.8±18.9
15
claude-opus-5-high1645.5±8.8
16
glm-5.3-max1642.2±13.3
17
qwen3.8-flash-next1638.4±14.6
18
mimo-v2.6-pro1627.9±18.7
19
mimo-v2.6-flash1627.5±18.3
20
glm-5.3-flash1626.1±10.4
21
gpt-5.6-sol-xhigh (codex-harness)1623.7±9.8
22
grok-4.6-high1623.3±12.6
23
muse-spark-1.3 (xHigh)1622.3±14.1
24
claude-fable-5-high1622.1±10.3
25
glm-5.2-max1620.4±10
26
deepseek-v4.1-flash-max1618.8±20
27
hy4-preview1615.2±18.5
28
gemini-4-argon-high1614.8±30.3
29
qwen3.8-27b1610.7±10.5
30
deepseek-v4-flash-high1606.8±17.5
31
step-5-preview-high1593.4±19.5
32
deepseek-v4-pro-high-202608131592.8±17.9
33
gpt-6-luna-max1582.4±10.1
34
claude-opus-4-7-high1564.2±9.1
35
claude-opus-4-71561.1±9
36
mistral-large-41559.4±25.7
37
claude-haiku-5.5-high1558.8±21.6
38
gemini-3.8-flash-high1557.1±12.9
39
claude-sonnet-5-high1555.9±10
40
grok-4.51554.6±11.4
41
gemini-3.7-flash-high1550.2±12.2
42
claude-opus-4-8-high1545.5±9.1
43
seed-2.1-pro-preview1537.3±10
44
muse-spark-1.11536.2±13.2
45
gpt-5.6-luna-xhigh (codex-harness)1535.9±11.1
46
qwen3.7-max-202605171534.7±12
47
claude-opus-4-81534.2±9.1
48
claude-opus-4-6-high1534±8.5
49
claude-opus-4-61532.5±8.4
50
gpt-5.6-terra-xhigh (codex-harness)1532.2±10.8
51
gemini-3.6-flash-high1526.5±13.4
52
kimi-k2.61524.2±11.5
53
claude-sonnet-4-61521.4±8.1
54
muse-spark-1.2 (xHigh)1521.4±25.4
55
hy31515.7±15.2
56
minimax-m31510.1±8.9
57
gpt-5.5-xhigh (codex-harness)1506.8±9.5
58
glm-5.11500±10.9
59
kimi-k2.7-code1497.2±16.6
60
gemini-3.5-flash-high1493.2±12.3
61
gemini-3.5-flash1488.9±28.1
62
gpt-5.5-high (codex-harness)1487.4±9.1
63
claude-opus-4-5-20251101-high-32k1484.6±12.6
64
qwen3.6-max-preview1479.1±20.9
65
gemini-3.5-flash-medium1471.9±11.1
66
mimo-v2.5-pro1469.4±8.3
67
claude-opus-4-5-202511011467.4±11.2
68
deepseek-v4-pro-high-preview1466.3±10
69
gpt-5.5 (codex-harness)1458.3±9.5
70
gemini-3-pro1458.2±12.4
71
gemini-3.1-pro-preview1458±7.9
72
deepseek-v4-pro1457.5±10
73
qwen3.6-plus1457.1±8.9
74
kimi-k2.5-thinking1454±8.3
75
inkling1453.3±10.2
76
gemini-3-flash1449.2±12.9
77
glm-4.71446.5±20.1
78
gpt-5.4-high (codex-harness)1445.7±32.4
79
Inkling Small1445.1±16.6
80
mimo-v2.51443.6±9.7
81
kimi-k2.5-instant1440.8±20.6
82
gpt-5.4-medium (codex-harness)1436.3±30.5
83
deepseek-v4-flash-high-preview1435.2±11.3
84
glm-51433.4±13.4
85
mimo-v2-pro1431.8±13.4
86
gpt-5-medium1431.3±25.6
87
minimax-m2.71406.7±9.3
88
minimax-m2.1-preview1399.1±14.4
89
qwen3.5-397b-a17b1399±8.1
90
glm-5v-turbo1398.3±23.5
91
gpt-5.1-medium1397.3±19.6
92
grok-4.20-beta-0309-reasoning1394±9.5
93
gpt-5.21393.4±41.4
94
gpt-5.4-mini-high1392.9±10.6
95
gemini-3-flash (thinking-minimal)1390.7±8
96
gpt-5.3-codex (codex-harness)1390±23.2
97
gemma-4-26b-a4b1389.8±27.4
98
solar-pro41386±19.3
99
claude-sonnet-4-5-20250929-high-32k1385.2±12
100
minimax-m2.51383.5±13.6
101
claude-sonnet-4-5-202509291383.2±11
102
muse-glimmer1382.2±29.8
103
gpt-5.41381.3±20.3
104
gemma-4-31b1381±10.3
105
qwen3.5-122b-a10b1377.6±12.5
106
grok-4.31370.7±9.9
107
claude-opus-4-1-202508051367.7±16.6
108
qwen3.5-27b1367.2±13
109
deepseek-v3.2-thinking1367.2±14.1
110
gpt-5.3-codex (codex-harness)1358.3±19
111
gpt-5.2-codex1354.4±14.1
112
gpt-5.11354.4±13.2
113
laguna-m.11352.1±16.3
114
gpt-5.1-codex1349.1±18.6
115
glm-4.61348.1±17.1
116
hunyuan-hy3-preview1344.7±29.8
117
mimo-v2-flash (non-thinking)1340.3±16.1
118
deepseek-v3.21334.7±12.8
119
kimi-k2-thinking-turbo1332.8±11.3
120
claude-haiku-4-5-202510011321.9±7.6
121
mimo-v2-flash (thinking)1316.9±27.6
122
laguna-xs.21308.9±18.6
123
minimax-m21307.8±17.3
124
qwen3-coder-480b-a35b-instruct1288.2±11.9
125
qwen3.5-flash1284±33.5
126
deepseek-v3.2-exp1270.3±22.9
127
mistral-medium-3.51269.4±24.9
128
qwen3.5-35b-a3b1260.3±30.7
129
grok-4-1-fast-reasoning1257.5±17.3
130
KAT-Coder-Pro-V11253.7±38.1
131
mistral-large-31251.9±43.5
132
gemini-3.1-flash-lite-preview1247.9±10.9
133
gpt-5.1-codex-mini1244.5±37.2
134
trinity-large-thinking1240.4±34.4
135
gemini-2.5-pro1229.4±27.2
136
devstral-21220.2±34.5
137
grok-4.1-thinking1211.1±46.9
138
granite-4.1-8b1203.9±30.8
139
grok-code-fast-11169.8±53.7
140
grok-4-fast-reasoning1158.1±53.6
141
mercury-21154.7±42.5
142
devstral-medium-25071114.6±49.9
Elo 分 真人在不知道模型名字的情况下两两投票，胜负累积出的等级分。绝对值没有意义，差值才有：分差 100 分 ≈ 强的那个有 64% 胜率，只差几分基本是平手。
± 后面的数 是 95% 置信区间的半宽，投票越多越小。两个模型的区间重叠，说明它们的名次差别在统计上不成立。
区间 把上面这个置信区间换算成名次：1–5 意思是这个模型的真实名次可能在第 1 到第 5 之间。
数据更新: 2026-10-14数据来源: Arena AI ↗