M
ModelSage
排行榜/Arena 人类盲测/Code 编程/游戏

Arena Code 编程 · 游戏

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-max1935.1±31.1
2
claude-sonnet-5.5-xhigh1895.2±26.5
3
gpt-6-astra-max1869.9±19.1
4
claude-fable-5.1-max1857.2±17
5
gpt-6.1-sol-max1842.6±25.1
6
claude-sonnet-5.5-high1831.9±18.5
7
gemini-4-argon-high1769.5±29
8
claude-opus-5-max1763.5±10.5
9
gpt-6-sol-max1755±19.8
10
qwen3.8-max-09021742.5±12
11
qwen3.8-max1737.2±22.9
12
muse-spark-1.3-max1729.7±13.4
13
claude-opus-5-high1726.2±9.2
14
kimi-k3-max1706±10.2
15
mimo-v2.6-flash1704.2±17.7
16
hy4-preview1698±17.3
17
grok-4.7-xhigh1695.3±18.4
18
claude-fable-5-high1695.2±10.5
19
qwen3.8-flash-next1694.5±14.1
20
muse-wonder-max1691.6±33.2
21
mimo-v2.6-pro1685.3±17.3
22
claude-haiku-5.5-high1681.9±23.8
23
muse-spark-1.3 (xHigh)1674.8±13.7
24
gpt-5.6-sol-xhigh (codex-harness)1673.9±10
25
grok-4.6-high1673±12.8
26
glm-5.3-max1671.4±13.3
27
deepseek-v4.1-flash-max1667.8±20.6
28
glm-5.3-flash1652.4±10.3
29
glm-5.2-max1646.1±10.4
30
gemini-3.7-flash-high1639±11.4
31
gemini-3.8-flash-high1635.8±12.5
32
qwen3.8-27b1635.7±10.4
33
step-5-preview-high1612.1±18
34
gpt-6-luna-max1611.9±9.8
35
claude-opus-4-8-high1606.7±9.9
36
deepseek-v4-pro-high-202608131605.7±17.6
37
deepseek-v4-flash-high1605.2±16.4
38
claude-opus-4-7-high1588.6±10
39
grok-4.51587.8±11.7
40
claude-opus-4-71587.6±9.6
41
claude-opus-4-6-high1584.2±9.9
42
mistral-large-41580.2±27.7
43
claude-sonnet-5-high1575.8±10.1
44
claude-opus-4-81570.7±9.5
45
claude-opus-4-61566.8±9.3
46
muse-spark-1.2 (xHigh)1561.3±26.7
47
gpt-5.5-xhigh (codex-harness)1558.9±10.4
48
gemini-3.6-flash-high1550.7±13.4
49
gpt-5.6-terra-xhigh (codex-harness)1547.4±11.1
50
muse-spark-1.11543.2±13.9
51
claude-sonnet-4-61541.3±9.1
52
hy31540.6±15.8
53
glm-5.11537.9±13.1
54
qwen3.7-max-202605171537.2±13.6
55
gpt-5.5-high (codex-harness)1530.1±10
56
seed-2.1-pro-preview1529.5±10.5
57
claude-opus-4-5-20251101-high-32k1528.2±15
58
gemini-3.5-flash1528.1±27.9
59
gemini-3.5-flash-high1527.2±12.9
60
gemini-3.5-flash-medium1525.8±11.4
61
gpt-5.6-luna-xhigh (codex-harness)1517±11.1
62
qwen3.6-max-preview1512.9±27.1
63
gpt-5.4-high (codex-harness)1505.2±41.3
64
mimo-v2.5-pro1499.4±8.8
65
kimi-k2.61498.3±14.1
66
gpt-5.5 (codex-harness)1487.8±10.2
67
claude-opus-4-5-202511011482.8±13.1
68
kimi-k2.7-code1478.5±17.2
69
minimax-m31473.7±9.5
70
deepseek-v4-pro-high-preview1472.1±11.4
71
qwen3.6-plus1468.1±9.8
72
gemini-3.1-pro-preview1467.9±8.9
73
gpt-5.4-medium (codex-harness)1467.4±41.8
74
gemini-3-pro1456.4±13.6
75
gemini-3-flash1453.5±14.4
76
deepseek-v4-pro1453±11.4
77
gpt-5.3-codex (codex-harness)1446.9±27.6
78
mimo-v2-pro1438.6±16.9
79
deepseek-v4-flash-high-preview1435.8±12.8
80
glm-51435.5±16.1
81
mimo-v2.51434.4±10.5
82
gpt-5.21433.2±45.1
83
gpt-5-medium1430.1±25.8
84
kimi-k2.5-thinking1427.5±9.3
85
glm-4.71426.1±21.3
86
gpt-5.41423.5±20.1
87
claude-opus-4-1-202508051421.9±17.1
88
gpt-5.4-mini-high1420.9±12.4
89
glm-5v-turbo1412.4±26.2
90
claude-sonnet-4-5-20250929-high-32k1403.9±12.9
91
gpt-5.3-codex (codex-harness)1403.8±24.8
92
gpt-5.1-medium1403.3±18.9
93
kimi-k2.5-instant1402±24.1
94
qwen3.5-397b-a17b1401.3±9.1
95
claude-sonnet-4-5-202509291392.7±12.2
96
inkling1392.5±10.8
97
gemini-3-flash (thinking-minimal)1391.8±9.1
98
minimax-m2.71383.6±10.5
99
Inkling Small1382.7±16.8
100
minimax-m2.51380.4±16.4
101
solar-pro41379.2±19.8
102
minimax-m2.1-preview1368±16.6
103
hunyuan-hy3-preview1360.8±36.8
104
grok-4.31355.9±10.9
105
grok-4.20-beta-0309-reasoning1355.5±11.4
106
gemma-4-31b1354.9±11.1
107
qwen3.5-122b-a10b1352±15.6
108
gpt-5.11348.3±14
109
gemma-4-26b-a4b1347±41.6
110
qwen3.5-27b1347±16.5
111
deepseek-v3.2-thinking1345.8±17.3
112
laguna-m.11334.9±17.8
113
gpt-5.1-codex1332.4±19.1
114
glm-4.61332.1±17.1
115
muse-glimmer1331.7±27.8
116
gpt-5.2-codex1327.2±17.5
117
mimo-v2-flash (non-thinking)1321.4±18.1
118
claude-haiku-4-5-202510011317.6±8.5
119
kimi-k2-thinking-turbo1311.7±13
120
deepseek-v3.21295.5±15
121
laguna-xs.21294±21.4
122
deepseek-v3.2-exp1278.6±21
123
minimax-m21276.1±17.1
124
mimo-v2-flash (thinking)1271.9±33.2
125
gemini-3.1-flash-lite-preview1262.9±12.5
126
qwen3-coder-480b-a35b-instruct1259.1±13.3
127
KAT-Coder-Pro-V11256.4±30.5
128
trinity-large-thinking1250.9±42.8
129
mistral-medium-3.51249.8±33.3
130
qwen3.5-35b-a3b1247.6±35.8
131
gpt-5.1-codex-mini1244.8±37.8
132
grok-4-1-fast-reasoning1242.3±20.4
133
gemini-2.5-pro1231.7±26.4
134
grok-4.1-thinking1229.5±42.6
135
mistral-large-31207.2±46.3
136
qwen3.5-flash1199.7±42.5
137
mercury-21198.9±54.3
138
granite-4.1-8b1181.7±39.9
139
devstral-21175.3±42
140
grok-4-fast-reasoning1159±51.7
141
grok-code-fast-11148.4±55.5
142
devstral-medium-25071035.5±62.9
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 ↗