M
ModelSage
排行榜/Arena 人类盲测/Code 编程/消费级产品

Arena Code 编程 · 消费级产品

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
gpt-6.1-sol-max1719.7±35.6
2
gpt-6-astra-max1701.3±24.8
3
claude-opus-5.5-max1699.2±30.3
4
claude-fable-5.1-max1659.4±19.2
5
gpt-6-sol-max1648.1±25.5
6
qwen3.8-max1636.2±31.5
7
claude-opus-5-max1631±12.2
8
claude-sonnet-5.5-xhigh1628.7±30.4
9
muse-wonder-max1626.3±50
10
qwen3.8-max-09021618.5±14.9
11
claude-sonnet-5.5-high1614.1±23.9
12
muse-spark-1.3-max1613±17.6
13
gemini-4-argon-high1611.3±35.8
14
kimi-k3-max1610±11.9
15
claude-opus-5-high1596.9±11
16
grok-4.6-high1588.2±16.9
17
deepseek-v4.1-flash-max1587±26.6
18
muse-spark-1.3 (xHigh)1582.2±17.9
19
grok-4.7-xhigh1579.8±23
20
hy4-preview1577.8±22.2
21
mimo-v2.6-flash1577.1±23.7
22
gpt-5.6-sol-xhigh (codex-harness)1575.1±11.5
23
glm-5.3-max1572.8±16.8
24
gemini-3.7-flash-high1571.4±15.5
25
claude-fable-5-high1570.5±12.1
26
mimo-v2.6-pro1566.5±24.3
27
glm-5.2-max1562.9±11.4
28
deepseek-v4-flash-high1562.8±21.1
29
deepseek-v4-pro-high-202608131561.2±21.1
30
qwen3.8-flash-next1559.6±18.9
31
qwen3.8-27b1557.7±13.3
32
glm-5.3-flash1554.6±12.7
33
mistral-large-41553.5±33.3
34
gpt-6-luna-max1553.4±12.4
35
claude-opus-4-7-high1541.2±9.8
36
claude-opus-4-8-high1540.3±10.3
37
gemini-3.8-flash-high1539±15.9
38
claude-opus-4-71537.2±10.1
39
muse-spark-1.2 (xHigh)1530.2±32.6
40
claude-opus-4-6-high1529.4±9.5
41
muse-spark-1.11529.3±15.3
42
claude-opus-4-61528.7±9.3
43
claude-opus-4-81528±10.6
44
grok-4.51524.8±14
45
claude-sonnet-4-61521.6±8.9
46
step-5-preview-high1520.7±26.5
47
claude-sonnet-5-high1518.7±12.3
48
gemini-3.6-flash-high1516.8±14.4
49
seed-2.1-pro-preview1509.4±12.3
50
qwen3.7-max-202605171508±12.9
51
kimi-k2.61507.7±12
52
gpt-5.6-terra-xhigh (codex-harness)1506.6±12.9
53
glm-5.11506.3±13.1
54
claude-haiku-5.5-high1505.5±29.1
55
gpt-5.6-luna-xhigh (codex-harness)1504.9±13
56
claude-opus-4-5-20251101-high-32k1496.3±14.4
57
minimax-m31481.6±10.4
58
hy31479.9±19.5
59
gpt-5.5-xhigh (codex-harness)1476.1±10.4
60
kimi-k2.7-code1471.8±17.9
61
gemini-3.5-flash-high1470.5±14.3
62
qwen3.6-max-preview1469.7±22.3
63
deepseek-v4-pro-high-preview1466.2±11.1
64
mimo-v2.5-pro1465.5±9.3
65
gemini-3.5-flash-medium1463.5±12.5
66
claude-opus-4-5-202511011462.5±12.5
67
gpt-5.5-high (codex-harness)1461.4±9.9
68
gpt-5.4-high (codex-harness)1460.9±33.5
69
qwen3.6-plus1452.6±9.5
70
mimo-v2.51451.7±10.6
71
deepseek-v4-pro1448.1±10.6
72
gemini-3-flash1446.8±14.7
73
gemini-3.5-flash1446.1±35.9
74
gpt-5.4-medium (codex-harness)1444.1±33.1
75
glm-51443.6±15.4
76
glm-5v-turbo1442.2±28.7
77
mimo-v2-pro1442.1±13.9
78
kimi-k2.5-thinking1436.6±9.3
79
gemini-3-pro1435.3±13.9
80
deepseek-v4-flash-high-preview1435.1±12.1
81
gpt-5.5 (codex-harness)1434.9±10.2
82
glm-4.71434.3±24.3
83
Inkling Small1424.7±21.2
84
inkling1423.4±13
85
gemini-3.1-pro-preview1422.6±8.4
86
gpt-5.21417.7±50.2
87
gpt-5-medium1416.4±33.4
88
minimax-m2.71405±10.6
89
gpt-5.3-codex (codex-harness)1401.4±25.4
90
qwen3.5-397b-a17b1400.7±8.9
91
gpt-5.1-medium1399.4±23.6
92
gpt-5.4-mini-high1393.1±11.4
93
claude-sonnet-4-5-20250929-high-32k1393±14
94
gpt-5.41386.5±21.5
95
claude-sonnet-4-5-202509291384.3±12.7
96
kimi-k2.5-instant1383.2±23.6
97
minimax-m2.1-preview1381.3±16.9
98
solar-pro41381.3±25
99
claude-opus-4-1-202508051378.1±20.2
100
gemini-3-flash (thinking-minimal)1373.7±8.8
101
minimax-m2.51370.6±14.6
102
laguna-m.11367.5±17.6
103
grok-4.20-beta-0309-reasoning1367±10
104
deepseek-v3.2-thinking1366.8±17
105
gpt-5.1-codex1362.8±23.4
106
gemma-4-31b1361.6±13.7
107
qwen3.5-27b1360.8±13.9
108
glm-4.61359.6±21.1
109
qwen3.5-122b-a10b1357.9±13.2
110
gpt-5.3-codex (codex-harness)1353.5±20.3
111
muse-glimmer1346.4±41
112
grok-4.31345.8±10.9
113
gpt-5.2-codex1340.8±16.3
114
gemma-4-26b-a4b1339±31.6
115
deepseek-v3.21338.6±14.4
116
gpt-5.11332.5±14.9
117
claude-haiku-4-5-202510011331±8.5
118
hunyuan-hy3-preview1324.9±31.9
119
mimo-v2-flash (non-thinking)1322.9±18.2
120
kimi-k2-thinking-turbo1319.6±13.1
121
laguna-xs.21317.7±20
122
minimax-m21309.9±20.9
123
mimo-v2-flash (thinking)1289.1±34.6
124
deepseek-v3.2-exp1281.4±29.5
125
mistral-medium-3.51280.4±26.6
126
qwen3-coder-480b-a35b-instruct1274.8±13.8
127
KAT-Coder-Pro-V11273.3±47.6
128
qwen3.5-35b-a3b1264.3±31.6
129
gemini-3.1-flash-lite-preview1254.7±11.2
130
grok-4-1-fast-reasoning1249.5±21
131
gpt-5.1-codex-mini1238.1±53
132
trinity-large-thinking1238±38.1
133
qwen3.5-flash1224.6±37.1
134
gemini-2.5-pro1221.4±34.3
135
grok-code-fast-11217.3±60.8
136
devstral-21197.7±43.4
137
granite-4.1-8b1185.9±33.9
138
grok-4.1-thinking1175.1±59.4
139
mistral-large-31172.6±64.4
140
mercury-21168.7±42.1
141
grok-4-fast-reasoning1155.8±57.7
142
devstral-medium-2507914.9±110
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 ↗