M
ModelSage
排行榜/Arena 人类盲测/Code 编程/Frontend 前端

Arena Code 编程 · Frontend 前端

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-max1854.2±17.7
2
claude-sonnet-5.5-xhigh1812.4±16
3
gpt-6-astra-max1803.2±12
4
claude-fable-5.1-max1777.4±10.5
5
gpt-6.1-sol-max1773.6±15.6
6
claude-sonnet-5.5-high1744.5±11.8
7
claude-opus-5-max1711.2±7
8
gpt-6-sol-max1690.2±12.4
9
gemini-4-argon-high1687.4±18.4
10
qwen3.8-max-09021685.6±7.9
11
qwen3.8-max1678.4±13.1
12
claude-opus-5-high1677.9±6.1
13
muse-spark-1.3-max1673.2±8.9
14
kimi-k3-max1670.1±6.8
15
grok-4.7-xhigh1660±11.5
16
muse-wonder-max1656.7±21.1
17
mimo-v2.6-flash1651.6±11.1
18
hy4-preview1651.1±10.9
19
qwen3.8-flash-next1650.7±9.1
20
mimo-v2.6-pro1639.9±11.6
21
claude-fable-5-high1637.2±6.8
22
glm-5.3-max1636.4±8.6
23
muse-spark-1.3 (xHigh)1634.9±9
24
grok-4.6-high1629.8±8.3
25
deepseek-v4.1-flash-max1629.7±12.4
26
gpt-5.6-sol-xhigh (codex-harness)1625.3±6.5
27
glm-5.3-flash1620.6±7.2
28
glm-5.2-max1613.6±6.7
29
claude-haiku-5.5-high1604.5±14
30
qwen3.8-27b1600.8±7.3
31
gemini-3.7-flash-high1592.4±7.6
32
deepseek-v4-pro-high-202608131589.5±10.7
33
gemini-3.8-flash-high1589.3±8.3
34
gpt-6-luna-max1587.6±6.8
35
deepseek-v4-flash-high1584.7±10.3
36
step-5-preview-high1577.6±11.8
37
grok-4.51562.5±7.3
38
claude-opus-4-7-high1562.4±6.1
39
claude-opus-4-8-high1560.5±6.3
40
claude-opus-4-71558.8±5.9
41
mistral-large-41549.9±15.9
42
claude-opus-4-6-high1548.2±5.7
43
claude-sonnet-5-high1543.8±6.6
44
claude-opus-4-81539.9±6.1
45
muse-spark-1.11539±8.5
46
claude-opus-4-61536.9±5.6
47
gemini-3.6-flash-high1536.2±8.5
48
muse-spark-1.2 (xHigh)1532.8±15.3
49
gpt-5.6-terra-xhigh (codex-harness)1522.6±7.1
50
claude-sonnet-4-61520.3±5.4
51
seed-2.1-pro-preview1517.9±6.6
52
qwen3.7-max-202605171517.8±7.8
53
gpt-5.5-xhigh (codex-harness)1516.8±6.3
54
gpt-5.6-luna-xhigh (codex-harness)1515.8±7.1
55
gemini-3.5-flash1513.8±17.6
56
hy31510.6±9.5
57
glm-5.11509.6±7.2
58
kimi-k2.61506.6±7.5
59
gemini-3.5-flash-high1503.2±7.9
60
claude-opus-4-5-20251101-high-32k1492.4±8.3
61
gemini-3.5-flash-medium1492±7.2
62
gpt-5.5-high (codex-harness)1490.7±6.3
63
qwen3.6-max-preview1482.5±12.6
64
minimax-m31479.8±5.9
65
mimo-v2.5-pro1475.6±5.6
66
kimi-k2.7-code1475.4±9.8
67
claude-opus-4-5-202511011468.2±7.2
68
gpt-5.4-high (codex-harness)1461.9±18.9
69
deepseek-v4-pro-high-preview1461.8±6.6
70
qwen3.6-plus1459.8±5.9
71
gpt-5.5 (codex-harness)1458.3±6.2
72
gemini-3.1-pro-preview1451.1±5.3
73
deepseek-v4-pro1445.2±6.7
74
gemini-3.5-flash-lite1442.7±48.7
75
gpt-5.4-medium (codex-harness)1442.5±18.3
76
gemini-3-flash1439±8.6
77
gemini-3-pro1438.8±8.4
78
kimi-k2.5-thinking1434.7±5.4
79
glm-51434±8.2
80
glm-4.71433.9±12.1
81
mimo-v2-pro1432.8±8.2
82
mimo-v2.51431±6.3
83
deepseek-v4-flash-high-preview1426.8±7.3
84
gpt-5-medium1415.8±15.6
85
gpt-5.21415.2±22.6
86
inkling1408±6.9
87
gpt-5.3-codex (codex-harness)1405.1±13.7
88
kimi-k2.5-instant1404.3±12.1
89
gpt-5.41403.5±11.6
90
Inkling Small1402.8±10.2
91
qwen3.5-397b-a17b1396.8±5.3
92
gpt-5.4-mini-high1395.9±6.8
93
gpt-5.1-medium1394.2±11.6
94
minimax-m2.71394.1±6
95
claude-sonnet-4-5-20250929-high-32k1392.2±7.7
96
claude-opus-4-1-202508051388.7±10.7
97
glm-5v-turbo1388.7±14.8
98
gemini-3-flash (thinking-minimal)1384.9±5.4
99
minimax-m2.51384.7±8.3
100
claude-sonnet-4-5-202509291383.6±7.3
101
minimax-m2.1-preview1382.8±9.6
102
grok-4.20-beta-0309-reasoning1374.2±6.3
103
solar-pro41369.7±11.5
104
gpt-5.3-codex (codex-harness)1369.6±11.4
105
gemma-4-31b1366.9±6.5
106
deepseek-v3.2-thinking1361.2±9.1
107
grok-4.31359.9±6.5
108
qwen3.5-122b-a10b1359.6±7.7
109
gemma-4-26b-a4b1357±17
110
qwen3.5-27b1357±8.1
111
hunyuan-hy3-preview1356.3±17.5
112
muse-glimmer1355.5±17
113
laguna-m.11343.3±9.8
114
gpt-5.11340.4±8.4
115
glm-4.61339.6±10.8
116
gpt-5.2-codex1337.3±8.9
117
gpt-5.1-codex1335.7±11.6
118
mimo-v2-flash (non-thinking)1329.2±9.8
119
claude-haiku-4-5-202510011326.3±5
120
deepseek-v3.21324.7±7.9
121
kimi-k2-thinking-turbo1320.6±7.4
122
laguna-xs.21298.8±11
123
minimax-m21295.6±10.9
124
mimo-v2-flash (thinking)1291.3±16.5
125
qwen3-coder-480b-a35b-instruct1273.4±7.8
126
deepseek-v3.2-exp1270.3±13.5
127
mistral-medium-3.51264.8±15.1
128
gemini-3.1-flash-lite-preview1254.8±6.9
129
KAT-Coder-Pro-V11254.2±19.5
130
qwen3.5-35b-a3b1252.8±17.5
131
qwen3.5-flash1242.5±19.6
132
gpt-5.1-codex-mini1242±22
133
grok-4-1-fast-reasoning1240.7±11.2
134
trinity-large-thinking1237.3±20.1
135
mistral-large-31228.9±25.5
136
gemini-2.5-pro1224.6±15.9
137
grok-4.1-thinking1212.9±24.9
138
devstral-21195.1±20.8
139
granite-4.1-8b1189.4±18.7
140
mercury-21169.4±24.5
141
grok-code-fast-11164.7±28.3
142
grok-4-fast-reasoning1157.3±27.4
143
devstral-medium-25071069.7±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 ↗