M
ModelSage
排行榜/Arena 人类盲测/Code 编程

Arena Code 编程排行榜

前端 / 全栈代码生成盲测，含 11 个细分场景子榜 · 完整榜单 143 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-max1811.1±14.6
2
gpt-6-astra-max1780.2±10.5
3
claude-sonnet-5.5-xhigh1773.7±13.9
4
gpt-6.1-sol-max1760.1±13.9
5
claude-fable-5.1-max1742.7±9.1
6
claude-sonnet-5.5-high1713±10.4
7
claude-opus-5-max1690±6.3
8
gpt-6-sol-max1685.9±11
9
gemini-4-argon-high1678±15.4
10
qwen3.8-max-09021673.2±7.3
11
qwen3.8-max1672±12
12
muse-spark-1.3-max1658.5±8.1
13
claude-opus-5-high1657.8±5.6
14
kimi-k3-max1655.3±6.1
15
muse-wonder-max1650.9±18.3
16
grok-4.7-xhigh1640.1±10.1
17
mimo-v2.6-flash1635.1±9.9
18
qwen3.8-flash-next1633.2±8.3
19
hy4-preview1631.6±9.6
20
mimo-v2.6-pro1625.2±10.3
21
claude-fable-5-high1623.8±6.3
22
muse-spark-1.3 (xHigh)1621.9±8.1
23
glm-5.3-max1621.2±7.8
24
deepseek-v4.1-flash-max1618.5±10.9
25
grok-4.6-high1617.4±7.5
26
gpt-5.6-sol-xhigh (codex-harness)1617±6
27
glm-5.3-flash1606.1±6.8
28
glm-5.2-max1604.1±6.1
29
qwen3.8-27b1594.2±6.7
30
gemini-3.7-flash-high1589.9±7
31
claude-haiku-5.5-high1585±12.2
32
gemini-3.8-flash-high1584.8±7.6
33
deepseek-v4-pro-high-202608131582.2±9.4
34
gpt-6-luna-max1581.7±6.3
35
deepseek-v4-flash-high1581.6±9.3
36
step-5-preview-high1563.7±10.5
37
claude-opus-4-7-high1557.6±5.8
38
claude-opus-4-8-high1556±5.8
39
claude-opus-4-71555.5±5.6
40
grok-4.51552.9±6.6
41
claude-opus-4-6-high1546.4±5.4
42
mistral-large-41542.3±13.8
43
muse-spark-1.11542.2±7.6
44
claude-sonnet-5-high1539.9±6
45
gemini-3.6-flash-high1538±7.5
46
claude-opus-4-61537.3±5.3
47
claude-opus-4-81536.2±5.7
48
muse-spark-1.2 (xHigh)1533.4±13.6
49
claude-sonnet-4-61522.3±5.1
50
gpt-5.6-terra-xhigh (codex-harness)1521.9±6.4
51
seed-2.1-pro-preview1519.7±6.1
52
gpt-5.6-luna-xhigh (codex-harness)1518.4±6.4
53
qwen3.7-max-202605171515.4±7.4
54
gpt-5.5-xhigh (codex-harness)1512.7±6
55
kimi-k2.61509.2±7.3
56
glm-5.11508.4±6.8
57
hy31506.8±8.5
58
gemini-3.5-flash1504.1±15.4
59
gemini-3.5-flash-high1499.4±7.3
60
claude-opus-4-5-20251101-high-32k1493.4±8.3
61
gemini-3.5-flash-medium1490.9±6.6
62
gpt-5.5-high (codex-harness)1487.4±5.9
63
minimax-m31482.1±5.5
64
qwen3.6-max-preview1482±12.5
65
mimo-v2.5-pro1480.5±5.3
66
kimi-k2.7-code1473.4±9.4
67
claude-opus-4-5-202511011469.2±7.1
68
deepseek-v4-pro-high-preview1463.8±6.4
69
gpt-5.4-high (codex-harness)1462.9±18.9
70
qwen3.6-plus1461.2±5.6
71
gpt-5.5 (codex-harness)1456.2±5.9
72
deepseek-v4-pro1446.8±6.3
73
gemini-3.1-pro-preview1446.7±5
74
gpt-5.4-medium (codex-harness)1443.6±18.3
75
gemini-3.5-flash-lite1440.4±40.6
76
gemini-3-flash1440±8.6
77
gemini-3-pro1439.9±8.5
78
mimo-v2.51438.2±6
79
kimi-k2.5-thinking1436.5±5.2
80
glm-51435.1±8.1
81
glm-4.71435±12
82
mimo-v2-pro1433.5±8.1
83
deepseek-v4-flash-high-preview1430.6±6.9
84
gpt-5-medium1417.2±15.6
85
gpt-5.21416.5±22.6
86
inkling1412.6±6.3
87
Inkling Small1408.4±9.1
88
gpt-5.3-codex (codex-harness)1406±13.6
89
kimi-k2.5-instant1405.3±12
90
gpt-5.41401.8±10.5
91
glm-5v-turbo1401±13.2
92
qwen3.5-397b-a17b1400.1±5.1
93
minimax-m2.71398.1±5.8
94
gpt-5.4-mini-high1397.6±6.5
95
gpt-5.1-medium1395.4±11.5
96
claude-sonnet-4-5-20250929-high-32k1393.3±7.7
97
claude-opus-4-1-202508051389.9±10.8
98
minimax-m2.51385.8±8.3
99
claude-sonnet-4-5-202509291384.7±7.3
100
gemini-3-flash (thinking-minimal)1384.6±5.2
101
minimax-m2.1-preview1383.8±9.6
102
grok-4.20-beta-0309-reasoning1375.3±6
103
gpt-5.3-codex (codex-harness)1370.9±11.4
104
solar-pro41370.7±10.3
105
gemma-4-31b1366.1±6.3
106
deepseek-v3.2-thinking1362.3±9
107
qwen3.5-122b-a10b1360.6±7.6
108
gemma-4-26b-a4b1359.2±16.9
109
qwen3.5-27b1357.9±8
110
grok-4.31357.3±6.3
111
hunyuan-hy3-preview1356.8±17.3
112
muse-glimmer1355.1±15.6
113
laguna-m.11348.5±9.3
114
gpt-5.11341.5±8.5
115
glm-4.61340.8±10.8
116
gpt-5.2-codex1338.4±8.9
117
gpt-5.1-codex1336.9±11.5
118
mimo-v2-flash (non-thinking)1330.2±9.8
119
claude-haiku-4-5-202510011329.9±4.9
120
deepseek-v3.21325.7±8
121
kimi-k2-thinking-turbo1321.8±7.5
122
laguna-xs.21302.6±10.7
123
minimax-m21296.9±10.9
124
mimo-v2-flash (thinking)1292.4±16.5
125
qwen3-coder-480b-a35b-instruct1274.6±7.8
126
deepseek-v3.2-exp1271.6±13.5
127
mistral-medium-3.51263.4±14.8
128
gemini-3.1-flash-lite-preview1256.3±6.8
129
KAT-Coder-Pro-V11255.6±19.5
130
qwen3.5-35b-a3b1253.8±17.5
131
qwen3.5-flash1243.5±19.6
132
gpt-5.1-codex-mini1243.4±22.1
133
grok-4-1-fast-reasoning1241.8±11.1
134
trinity-large-thinking1237.6±20
135
mistral-large-31230.2±25.5
136
gemini-2.5-pro1226±15.9
137
grok-4.1-thinking1214.3±24.9
138
devstral-21196.2±20.8
139
granite-4.1-8b1191.4±18.4
140
mercury-21170.6±24.5
141
grok-code-fast-11166.1±28.3
142
grok-4-fast-reasoning1158.7±27.4
143
devstral-medium-25071071.2±30.6
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 ↗