M
ModelSage
排行榜/Arena 人类盲测/Code 编程/参考图还原

Arena Code 编程 · 参考图还原

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-max1845.7±38.5
2
claude-sonnet-5.5-xhigh1831.1±32.5
3
gpt-6-astra-max1815.2±24.2
4
claude-fable-5.1-max1796±20.1
5
gpt-6.1-sol-max1787.8±31.3
6
claude-sonnet-5.5-high1768.8±24
7
gpt-6-sol-max1726.8±24.9
8
claude-opus-5-max1709.1±12.1
9
gemini-4-argon-high1703.9±36.9
10
qwen3.8-max1684.3±28.3
11
qwen3.8-max-09021683.6±14
12
claude-opus-5-high1682.5±10.7
13
kimi-k3-max1676.6±12
14
muse-spark-1.3-max1668.5±16.3
15
muse-wonder-max1661.1±43
16
gpt-5.6-sol-xhigh (codex-harness)1655±12
17
claude-fable-5-high1653.5±12
18
mimo-v2.6-flash1651.5±21
19
hy4-preview1649.4±21.7
20
grok-4.7-xhigh1648.2±22.1
21
mimo-v2.6-pro1641±22.4
22
glm-5.3-max1639.4±15.9
23
grok-4.6-high1638.5±16
24
qwen3.8-flash-next1635.8±17.5
25
muse-spark-1.3 (xHigh)1623.4±17.2
26
glm-5.3-flash1614.9±12.1
27
deepseek-v4.1-flash-max1613.2±24.1
28
glm-5.2-max1610.2±11.6
29
claude-haiku-5.5-high1603.4±27.3
30
gemini-3.7-flash-high1603.2±14.1
31
qwen3.8-27b1598.3±12.4
32
gemini-3.8-flash-high1594.9±14.9
33
gpt-6-luna-max1594.3±12.1
34
deepseek-v4-pro-high-202608131583.4±21
35
claude-opus-4-7-high1580.5±10
36
deepseek-v4-flash-high1577.1±20.4
37
claude-opus-4-8-high1576.8±10.5
38
grok-4.51575.8±14.4
39
step-5-preview-high1573.6±23.5
40
claude-opus-4-71571.4±9.9
41
claude-opus-4-6-high1560.3±9.7
42
mistral-large-41558.3±32.3
43
muse-spark-1.11557.9±16.9
44
claude-opus-4-81554.4±10.4
45
muse-spark-1.2 (xHigh)1552.7±30.8
46
claude-sonnet-5-high1551.3±12.2
47
claude-opus-4-61550.4±9.3
48
gemini-3.6-flash-high1540.5±16.1
49
qwen3.7-max-202605171537.6±13.3
50
glm-5.11535.1±13.5
51
gpt-5.5-xhigh (codex-harness)1534.4±10.4
52
claude-sonnet-4-61530.1±8.8
53
seed-2.1-pro-preview1524.9±11.9
54
gpt-5.6-terra-xhigh (codex-harness)1520±13.4
55
gpt-5.6-luna-xhigh (codex-harness)1518.1±13.1
56
claude-opus-4-5-20251101-high-32k1518±14
57
hy31511.5±19.6
58
kimi-k2.61509.5±12.6
59
gemini-3.5-flash1506±33.5
60
gemini-3.5-flash-high1502.9±14.2
61
gpt-5.5-high (codex-harness)1498.4±10.1
62
gemini-3.5-flash-medium1492.9±13.3
63
kimi-k2.7-code1482±17.7
64
mimo-v2.5-pro1477.9±9.1
65
minimax-m31477.2±10.1
66
gpt-5.4-high (codex-harness)1476.5±34.1
67
deepseek-v4-pro-high-preview1474.8±11.2
68
claude-opus-4-5-202511011472.7±12.2
69
gpt-5.4-medium (codex-harness)1471.1±33.2
70
qwen3.6-max-preview1470.2±22.6
71
gpt-5.5 (codex-harness)1466.2±10.3
72
qwen3.6-plus1466±9.7
73
gemini-3.1-pro-preview1455.2±8.6
74
deepseek-v4-pro1453.9±11.1
75
gemini-3-pro1450.8±13
76
mimo-v2.51446.9±10.4
77
mimo-v2-pro1444.2±14.3
78
glm-51443.5±15.4
79
gemini-3-flash1442.5±14
80
deepseek-v4-flash-high-preview1442.1±12.5
81
kimi-k2.5-thinking1440.4±9.1
82
glm-4.71436.5±21.5
83
gpt-5.41420.6±22.4
84
gpt-5-medium1419.5±24.3
85
kimi-k2.5-instant1417.5±23.7
86
gpt-5.3-codex (codex-harness)1415.9±25.8
87
claude-opus-4-1-202508051413.8±16.8
88
gpt-5.4-mini-high1410.7±11.5
89
inkling1402.2±13
90
qwen3.5-397b-a17b1399±9
91
Inkling Small1396.1±20.6
92
gpt-5.21393.9±43
93
gpt-5.1-medium1393.7±19.4
94
claude-sonnet-4-5-20250929-high-32k1393.4±12.4
95
minimax-m2.71391.1±10.6
96
glm-5v-turbo1388.5±32.4
97
gemini-3-flash (thinking-minimal)1387.8±8.6
98
claude-sonnet-4-5-202509291386.8±11.7
99
gpt-5.3-codex (codex-harness)1384.8±20.6
100
grok-4.20-beta-0309-reasoning1380.5±10.4
101
minimax-m2.1-preview1373±15.9
102
muse-glimmer1368.1±35.1
103
minimax-m2.51367.9±15
104
solar-pro41367.2±25.1
105
qwen3.5-122b-a10b1363.3±13.6
106
qwen3.5-27b1360.8±14.1
107
grok-4.31359.9±11.1
108
gemma-4-26b-a4b1359.5±32.8
109
hunyuan-hy3-preview1355±31.6
110
deepseek-v3.2-thinking1352.9±15.7
111
laguna-m.11348.3±18
112
gemma-4-31b1346.9±12.6
113
gpt-5.2-codex1346.3±15.9
114
gpt-5.11344.3±13.4
115
glm-4.61338.3±16.9
116
gpt-5.1-codex1332.8±19.1
117
mimo-v2-flash (non-thinking)1319.9±17.4
118
deepseek-v3.21319.1±13.8
119
claude-haiku-4-5-202510011318.2±8.4
120
kimi-k2-thinking-turbo1313.4±12.1
121
laguna-xs.21289.8±20
122
minimax-m21289.6±16.8
123
mistral-medium-3.51276.9±25.7
124
deepseek-v3.2-exp1273.3±20.6
125
qwen3-coder-480b-a35b-instruct1264.3±12.6
126
mimo-v2-flash (thinking)1262.6±31.4
127
gemini-3.1-flash-lite-preview1259.5±11.5
128
KAT-Coder-Pro-V11257.8±33.3
129
qwen3.5-35b-a3b1250.7±31.5
130
qwen3.5-flash1240.6±35.2
131
grok-4-1-fast-reasoning1235.4±19.3
132
gpt-5.1-codex-mini1232.1±36.9
133
trinity-large-thinking1228.4±39.9
134
gemini-2.5-pro1225.9±25.7
135
grok-4.1-thinking1217.1±45.4
136
granite-4.1-8b1216.6±30.9
137
mistral-large-31209.9±48.1
138
mercury-21184.3±46.3
139
devstral-21181.4±39.4
140
grok-code-fast-11177.7±51.5
141
grok-4-fast-reasoning1148.8±53.1
142
devstral-medium-25071081±55.7
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 ↗