M
ModelSage
排行榜/Arena 人类盲测/Code 编程/仿真模拟

Arena Code 编程 · 仿真模拟

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-max1893.5±49.4
2
gpt-6-astra-max1889.4±26.8
3
claude-sonnet-5.5-xhigh1842.9±40.4
4
claude-fable-5.1-max1823.7±24.1
5
gpt-6.1-sol-max1822.6±34.6
6
claude-sonnet-5.5-high1807.5±27
7
claude-opus-5-max1750.6±14.9
8
gemini-4-argon-high1747.3±45.7
9
gpt-6-sol-max1735.8±26.3
10
claude-opus-5-high1723±12.3
11
hy4-preview1703.5±24.6
12
muse-wonder-max1700.4±50
13
mimo-v2.6-pro1700.3±26
14
qwen3.8-max-09021699.1±16
15
muse-spark-1.3-max1692.1±18.2
16
grok-4.7-xhigh1688.1±25.1
17
kimi-k3-max1684.2±13.5
18
qwen3.8-max1676.3±30.3
19
claude-fable-5-high1673.2±14.9
20
mimo-v2.6-flash1661.5±24.8
21
glm-5.3-max1656.3±18.2
22
qwen3.8-flash-next1653.8±18.7
23
grok-4.6-high1653±16.9
24
muse-spark-1.3 (xHigh)1634.5±19
25
glm-5.2-max1633.1±14.7
26
deepseek-v4.1-flash-max1632.3±27.8
27
gpt-5.6-sol-xhigh (codex-harness)1627.5±13.5
28
claude-haiku-5.5-high1624.3±34.5
29
glm-5.3-flash1621.6±13.5
30
gemini-3.7-flash-high1619.5±15.5
31
deepseek-v4-pro-high-202608131618.9±26.5
32
gpt-6-luna-max1608.3±13.1
33
gemini-3.8-flash-high1607.1±16.9
34
qwen3.8-27b1598.9±14.3
35
claude-opus-4-8-high1576.5±13.3
36
grok-4.51574.3±16
37
claude-opus-4-7-high1573.4±13.4
38
step-5-preview-high1571.2±25.6
39
deepseek-v4-flash-high1568.8±24.1
40
claude-opus-4-71568.3±13.1
41
claude-opus-4-6-high1563.4±13.1
42
mistral-large-41559.2±38.6
43
gemini-3.6-flash-high1554.5±19.8
44
claude-sonnet-5-high1554.4±14.3
45
gpt-5.5-xhigh (codex-harness)1548.1±14.4
46
claude-opus-4-61546.7±12.5
47
gemini-3.5-flash1546±39.4
48
muse-spark-1.11544.4±19.4
49
glm-5.11542.5±17.9
50
claude-opus-4-81538.3±13
51
gpt-5.6-terra-xhigh (codex-harness)1531.7±14.9
52
hy31527±21.8
53
qwen3.6-max-preview1524.8±34.4
54
qwen3.7-max-202605171522.8±18.9
55
gemini-3.5-flash-high1522.3±18.4
56
claude-sonnet-4-61518±11.9
57
muse-spark-1.2 (xHigh)1516.3±34.1
58
gpt-5.5-high (codex-harness)1509.8±13.7
59
gemini-3.5-flash-medium1505.9±16.5
60
seed-2.1-pro-preview1503±14.3
61
gpt-5.6-luna-xhigh (codex-harness)1502.3±15.1
62
kimi-k2.61500.2±18.3
63
claude-opus-4-5-20251101-high-32k1500±18.4
64
mimo-v2.5-pro1490.4±12.1
65
gpt-5.4-high (codex-harness)1479.7±58.1
66
gpt-5.5 (codex-harness)1476.4±14.1
67
gemini-3.1-pro-preview1473.4±11.9
68
gpt-5.4-medium (codex-harness)1472.3±58.5
69
qwen3.6-plus1470.6±13.4
70
claude-opus-4-5-202511011467±16.5
71
kimi-k2.7-code1465.4±23.9
72
deepseek-v4-pro-high-preview1465.1±15
73
minimax-m31461.8±12.8
74
gemini-3-pro1453.5±16.4
75
deepseek-v4-pro1442±15
76
glm-51440.4±20.5
77
mimo-v2-pro1434.9±21.4
78
kimi-k2.5-thinking1427.8±13
79
gpt-5.41426.7±26.9
80
minimax-m2.51424.9±20.1
81
gemini-3-flash1423.8±17.4
82
gpt-5.21418.2±50.6
83
mimo-v2.51417±14.6
84
deepseek-v4-flash-high-preview1412.9±17.9
85
gpt-5.3-codex (codex-harness)1412.7±39.5
86
gpt-5.4-mini-high1406.2±16.5
87
glm-4.71402.6±25.2
88
claude-sonnet-4-5-20250929-high-32k1393.3±16.1
89
Inkling Small1392.7±24.1
90
gpt-5.3-codex (codex-harness)1388.8±32
91
kimi-k2.5-instant1386.4±30.6
92
minimax-m2.71384.8±14
93
qwen3.5-397b-a17b1383.6±12.4
94
claude-sonnet-4-5-202509291381.2±15.3
95
minimax-m2.1-preview1381.2±19.4
96
inkling1381±14
97
gpt-5-medium1380.1±31.2
98
gemini-3-flash (thinking-minimal)1377.7±11.8
99
solar-pro41376.8±25.3
100
claude-opus-4-1-202508051376.1±21.4
101
gpt-5.1-medium1375.8±24.1
102
muse-glimmer1368.9±38.1
103
hunyuan-hy3-preview1366±48.4
104
glm-5v-turbo1364.1±35.1
105
gemma-4-31b1362±14.7
106
grok-4.31358±14.9
107
grok-4.20-beta-0309-reasoning1349.3±15.1
108
deepseek-v3.2-thinking1346.1±21
109
qwen3.5-27b1332.1±20.6
110
gpt-5.2-codex1331.1±21.3
111
laguna-m.11328.4±25.1
112
gemma-4-26b-a4b1328.3±50.9
113
qwen3.5-122b-a10b1328.1±20.5
114
gpt-5.1-codex1326.8±24.2
115
mimo-v2-flash (non-thinking)1324.5±22
116
gpt-5.11324±17.6
117
glm-4.61320.2±21.3
118
claude-haiku-4-5-202510011310.4±10.9
119
kimi-k2-thinking-turbo1301.1±15.9
120
deepseek-v3.21296.5±18.6
121
mistral-medium-3.51292.1±38.8
122
laguna-xs.21291.8±28.8
123
mimo-v2-flash (thinking)1274±41.9
124
qwen3-coder-480b-a35b-instruct1267±16.6
125
minimax-m21259.7±21.3
126
deepseek-v3.2-exp1258.8±27.3
127
gemini-3.1-flash-lite-preview1258.3±16.8
128
mistral-large-31251.2±61
129
grok-4-1-fast-reasoning1248.6±26.1
130
KAT-Coder-Pro-V11245.8±46.2
131
trinity-large-thinking1241.7±54.3
132
qwen3.5-35b-a3b1237.4±47.6
133
gpt-5.1-codex-mini1229.7±46.3
134
mercury-21211.4±71.3
135
gemini-2.5-pro1206.3±33
136
qwen3.5-flash1183±63.6
137
granite-4.1-8b1166.7±45.8
138
devstral-21143.7±53.4
139
grok-code-fast-11141.4±67.8
140
grok-4.1-thinking1134±61.8
141
grok-4-fast-reasoning1128.6±64.6
142
devstral-medium-25071037.8±69.6
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 ↗