M
ModelSage
排行榜/Arena 人类盲测/Code 编程/React

Arena Code 编程 · React

前端 / 全栈代码生成盲测,含 11 个细分场景子榜 · 完整榜单 127 个模型

RankModelElo 分
1
claude-opus-5.5-max1859±18.4
2
claude-sonnet-5.5-xhigh1823.9±17.6
3
gpt-6-astra-max1804.4±13.5
4
claude-fable-5.1-max1791±11.4
5
gpt-6.1-sol-max1760.8±17.3
6
claude-sonnet-5.5-high1748±13
7
claude-opus-5-max1712.3±7.5
8
gemini-4-argon-high1690.5±20
9
qwen3.8-max-09021690±8.5
10
gpt-6-sol-max1688.3±14
11
qwen3.8-max1682±14.2
12
claude-opus-5-high1677.3±6.5
13
kimi-k3-max1670.4±7.1
14
muse-spark-1.3-max1670.2±9.5
15
grok-4.7-xhigh1663.3±12.7
16
qwen3.8-flash-next1657.7±10.1
17
hy4-preview1652.8±11.9
18
mimo-v2.6-flash1647.6±12.1
19
glm-5.3-max1640±9.3
20
mimo-v2.6-pro1639.5±12.5
21
muse-wonder-max1637.8±23.2
22
grok-4.6-high1631.2±8.9
23
muse-spark-1.3 (xHigh)1630.9±9.8
24
claude-fable-5-high1630.7±7.2
25
deepseek-v4.1-flash-max1628.1±13.9
26
glm-5.3-flash1625±7.6
27
gpt-5.6-sol-xhigh (codex-harness)1621.8±7.1
28
glm-5.2-max1616.6±7.1
29
claude-haiku-5.5-high1606.9±15.4
30
qwen3.8-27b1603±7.8
31
deepseek-v4-pro-high-202608131588±11.6
32
gemini-3.7-flash-high1586.6±8.3
33
deepseek-v4-flash-high1584.4±11.2
34
step-5-preview-high1579.3±12.9
35
gpt-6-luna-max1576.9±7.3
36
gemini-3.8-flash-high1576.6±8.9
37
claude-opus-4-7-high1560.2±6.4
38
claude-opus-4-8-high1556.8±6.6
39
mistral-large-41555.9±17.3
40
grok-4.51555.8±7.8
41
claude-opus-4-71552.7±6.3
42
claude-sonnet-5-high1540.8±7
43
claude-opus-4-6-high1538.9±6.1
44
claude-opus-4-81537.5±6.5
45
muse-spark-1.11531.8±9.1
46
claude-opus-4-61528.4±5.9
47
gemini-3.6-flash-high1527.6±9.1
48
muse-spark-1.2 (xHigh)1525.3±16.5
49
seed-2.1-pro-preview1516.6±6.9
50
qwen3.7-max-202605171513.8±8.3
51
gpt-5.6-terra-xhigh (codex-harness)1513.3±7.6
52
claude-sonnet-4-61511.9±5.6
53
hy31510.8±10.5
54
gpt-5.6-luna-xhigh (codex-harness)1507.9±7.8
55
gpt-5.5-xhigh (codex-harness)1506.2±6.7
56
glm-5.11505.2±7.7
57
kimi-k2.61505±8
58
gemini-3.5-flash1496.3±19.9
59
gemini-3.5-flash-high1494.9±8.4
60
gemini-3.5-flash-medium1481.4±7.6
61
minimax-m31478±6.3
62
claude-opus-4-5-20251101-high-32k1477.9±10.6
63
gpt-5.5-high (codex-harness)1477.7±6.6
64
kimi-k2.7-code1474.6±10.5
65
qwen3.6-max-preview1473.4±13.5
66
mimo-v2.5-pro1469.5±5.9
67
claude-opus-4-5-202511011456.7±9
68
deepseek-v4-pro-high-preview1453.7±6.9
69
qwen3.6-plus1453.4±6.1
70
gpt-5.4-high (codex-harness)1452.5±19.8
71
gpt-5.5 (codex-harness)1446±6.6
72
deepseek-v4-pro1440.3±7
73
gemini-3.1-pro-preview1438.2±5.6
74
gpt-5.4-medium (codex-harness)1429±19.4
75
kimi-k2.5-thinking1428.9±5.7
76
gemini-3-flash1428±12
77
mimo-v2.51424.7±6.7
78
mimo-v2-pro1424.7±8.7
79
glm-51423.9±8.8
80
deepseek-v4-flash-high-preview1421.6±7.7
81
glm-4.71419.7±67.3
82
gemini-3.5-flash-lite1417.8±52
83
inkling1402.7±7.4
84
Inkling Small1397.9±11
85
gemini-3-pro1395.3±12.5
86
gpt-5.3-codex (codex-harness)1394.4±14.6
87
kimi-k2.5-instant1391.8±13.2
88
gpt-5.41391.4±12.8
89
minimax-m2.71387.5±6.4
90
qwen3.5-397b-a17b1387.3±5.7
91
gpt-5.4-mini-high1386.5±7.1
92
glm-5v-turbo1385.7±15.7
93
claude-sonnet-4-5-20250929-high-32k1385.2±11
94
claude-sonnet-4-5-202509291382.1±9.8
95
gemini-3-flash (thinking-minimal)1374.4±6
96
minimax-m2.51373.4±8.9
97
grok-4.20-beta-0309-reasoning1368.2±6.6
98
minimax-m2.1-preview1364.3±14.5
99
solar-pro41364±13
100
gemma-4-31b1358.3±7
101
gpt-5.3-codex (codex-harness)1356.2±12
102
grok-4.31352.8±6.9
103
qwen3.5-122b-a10b1352.6±8.2
104
gemma-4-26b-a4b1350.5±18.4
105
deepseek-v3.2-thinking1349.7±11.6
106
hunyuan-hy3-preview1348.8±19.1
107
muse-glimmer1348.3±18.4
108
qwen3.5-27b1345.6±8.6
109
laguna-m.11338.2±10.7
110
deepseek-v3.21333.4±9.9
111
gpt-5.2-codex1329±10.5
112
claude-haiku-4-5-202510011325.8±6
113
kimi-k2-thinking-turbo1313.8±9.9
114
mimo-v2-flash (non-thinking)1311.3±14.2
115
gpt-5.11306.4±13.7
116
laguna-xs.21294.8±11.8
117
qwen3-coder-480b-a35b-instruct1257.9±11.1
118
mistral-medium-3.51253.4±16.2
119
gemini-3.1-flash-lite-preview1244.8±7.4
120
qwen3.5-35b-a3b1237.8±18.8
121
trinity-large-thinking1230.5±21.9
122
qwen3.5-flash1229.1±20.6
123
mimo-v2-flash (thinking)1228.5±24
124
grok-4-1-fast-reasoning1221.4±20.2
125
granite-4.1-8b1180.3±20.3
126
mercury-21159.8±26.3
127
devstral-21146.3±50.4
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 ↗