M
ModelSage
排行榜/Arena 人类盲测/Code 编程/数据分析应用

Arena Code 编程 · 数据分析应用

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-max1694.7±29.1
2
gpt-6-astra-max1685±22.4
3
gpt-6.1-sol-max1653.5±33.2
4
gemini-4-argon-high1632.9±32.4
5
gpt-6-sol-max1623.8±24
6
claude-sonnet-5.5-xhigh1618.3±28.6
7
qwen3.8-max1614.4±28.5
8
qwen3.8-max-09021610.3±13.9
9
claude-fable-5.1-max1606.7±17.3
10
claude-opus-5-max1601.6±10.9
11
claude-sonnet-5.5-high1593.1±22.4
12
muse-spark-1.3-max1588.4±16.6
13
kimi-k3-max1579.9±11.1
14
qwen3.8-flash-next1579.2±16.8
15
claude-opus-5-high1579.1±10.1
16
gemini-3.7-flash-high1578.4±14
17
muse-wonder-max1576±44
18
gpt-5.6-sol-xhigh (codex-harness)1568.8±10.8
19
muse-spark-1.3 (xHigh)1566.2±16.8
20
glm-5.3-max1564.4±15
21
grok-4.6-high1562.6±15.2
22
deepseek-v4.1-flash-max1559±24.6
23
grok-4.7-xhigh1557.6±21.6
24
mimo-v2.6-flash1557.1±21.1
25
hy4-preview1556.2±20.8
26
gemini-3.8-flash-high1554.8±14.8
27
mimo-v2.6-pro1554.5±23
28
claude-fable-5-high1553.9±11.4
29
glm-5.3-flash1551.7±11.8
30
qwen3.8-27b1550±12.4
31
glm-5.2-max1546.9±10.7
32
gpt-6-luna-max1541.9±11.5
33
deepseek-v4-flash-high1537.4±20.4
34
muse-spark-1.11537.4±14.9
35
deepseek-v4-pro-high-202608131533.8±19.8
36
claude-opus-4-7-high1528.4±9.4
37
claude-opus-4-71522.3±9.6
38
grok-4.51521.3±13.1
39
gemini-3.6-flash-high1521.1±13.5
40
claude-opus-4-8-high1520.2±9.8
41
claude-sonnet-4-61518.4±8.6
42
claude-opus-4-6-high1518.3±9.1
43
muse-spark-1.2 (xHigh)1518±30.8
44
claude-opus-4-61514.3±9
45
gpt-5.6-luna-xhigh (codex-harness)1511.3±12.3
46
step-5-preview-high1507.1±23
47
seed-2.1-pro-preview1503.3±11.6
48
claude-sonnet-5-high1502.9±11.6
49
claude-haiku-5.5-high1501.8±25.5
50
mistral-large-41500.3±29.4
51
claude-opus-4-81500.2±10.1
52
qwen3.7-max-202605171500±12.8
53
glm-5.11497.4±12.1
54
claude-opus-4-5-20251101-high-32k1495.6±14.3
55
gpt-5.6-terra-xhigh (codex-harness)1494.7±12
56
kimi-k2.61491.5±11.8
57
minimax-m31485.5±9.6
58
qwen3.6-max-preview1477.9±22.4
59
gemini-3.5-flash-high1473.6±13.5
60
gemini-3.5-flash-medium1470.1±11.8
61
gemini-3.5-flash1469±33.1
62
gpt-5.5-xhigh (codex-harness)1467.3±10
63
mimo-v2.5-pro1466.1±8.9
64
hy31466.1±18.1
65
kimi-k2.7-code1456.7±17.8
66
deepseek-v4-pro-high-preview1453.6±10.8
67
qwen3.6-plus1453.1±9
68
gpt-5.5-high (codex-harness)1451.2±9.6
69
claude-opus-4-5-202511011448.7±12.4
70
mimo-v2.51442.1±10.3
71
deepseek-v4-pro1437.2±10.1
72
gpt-5.4-high (codex-harness)1436.9±34.7
73
mimo-v2-pro1431.8±13.4
74
glm-51431.3±15.6
75
gpt-5.5 (codex-harness)1425.4±9.9
76
deepseek-v4-flash-high-preview1424.5±11.9
77
kimi-k2.5-thinking1423.8±9.2
78
gpt-5.4-medium (codex-harness)1423.2±32.7
79
glm-4.71420±24.5
80
gpt-5.21419.4±47.9
81
gemini-3-pro1417.8±13.9
82
gemini-3-flash1415.3±14.5
83
Inkling Small1412.6±19.9
84
inkling1407.2±12.4
85
gemini-3.1-pro-preview1406±8.3
86
minimax-m2.71401.2±10.3
87
qwen3.5-397b-a17b1400.5±8.6
88
gpt-5.41395.8±20.2
89
claude-sonnet-4-5-20250929-high-32k1395.2±13.9
90
gpt-5.3-codex (codex-harness)1395±24.5
91
glm-5v-turbo1393.7±27
92
gpt-5.1-medium1391.9±23.6
93
gpt-5.4-mini-high1388.7±11.1
94
minimax-m2.51387.7±14.4
95
claude-sonnet-4-5-202509291385.1±12.4
96
deepseek-v3.2-thinking1382.7±17.8
97
minimax-m2.1-preview1381.4±17.3
98
gpt-5-medium1378.5±34.6
99
claude-opus-4-1-202508051367.4±21.1
100
gemini-3-flash (thinking-minimal)1364.4±8.6
101
gpt-5.1-codex1362.1±24
102
grok-4.20-beta-0309-reasoning1361.6±9.7
103
gpt-5.3-codex (codex-harness)1360.6±20
104
laguna-m.11359.8±17.3
105
kimi-k2.5-instant1359.6±23.3
106
solar-pro41356.1±23.6
107
qwen3.5-27b1353.5±13.9
108
gpt-5.11351.9±15.3
109
qwen3.5-122b-a10b1350.9±13.2
110
gpt-5.2-codex1344.9±16.1
111
glm-4.61343.6±21.3
112
claude-haiku-4-5-202510011340.2±8.3
113
gemma-4-26b-a4b1337.1±32
114
gemma-4-31b1335.9±14.1
115
grok-4.31335.6±10.6
116
hunyuan-hy3-preview1331.2±30.8
117
kimi-k2-thinking-turbo1321.6±13.3
118
muse-glimmer1320.7±42.8
119
mimo-v2-flash (non-thinking)1319.2±18.6
120
deepseek-v3.21317.7±14.4
121
minimax-m21314±20.7
122
laguna-xs.21303.3±19.3
123
mimo-v2-flash (thinking)1297.8±32
124
KAT-Coder-Pro-V11282.3±48.9
125
qwen3-coder-480b-a35b-instruct1274.9±13.9
126
mistral-large-31268.3±55
127
deepseek-v3.2-exp1260.8±29.5
128
qwen3.5-35b-a3b1256.2±31.9
129
mistral-medium-3.51252.9±26.1
130
grok-4.1-thinking1250.2±60.3
131
gemini-3.1-flash-lite-preview1238.3±11
132
trinity-large-thinking1236.2±33.5
133
grok-4-1-fast-reasoning1235.1±20.6
134
qwen3.5-flash1234.5±33.4
135
grok-code-fast-11214.7±63.9
136
devstral-21213.7±41.2
137
gpt-5.1-codex-mini1212.4±56.6
138
grok-4-fast-reasoning1194±56.7
139
granite-4.1-8b1171.2±35.1
140
gemini-2.5-pro1163.8±39.1
141
mercury-21144.1±45.5
142
devstral-medium-25071012.8±93.9
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 ↗