M
ModelSage
排行榜/Arena 人类偏好

Arena AI 人类偏好排行榜

基于Arena AI Text Elo评分,反映真实用户投票偏好。

分数怎么算:分数 = (Elo − 900) ÷ (全站最高 Elo − 900) × 100。所有 Arena 榜(文本 / 代码 / 视觉 / 搜索 / 文档)共用这一把尺子,所以不同榜的分数可以横着比,也因此单个榜的榜首通常不到 100 分。

Arena AI 数据更新于 2026-10-14,每天北京时间凌晨自动刷新

RankModelScore
1
Gemini 4 Argon (high)
68.6
2
Claude Opus 5.5 (high with fallback)
66.7
3
Claude Fable 5.1 (max with fallback)
66.2
4
Claude Opus 4.6 (max)
66.2
5
Claude Fable 5.1 (high with fallback)
65.9
6
Claude Opus 4.6 (non-reasoning, high)
65.6
7
Gemini 3.8 Flash (high)
65.4
8
Claude Opus 4.7 (max)
65.2
9
Claude Opus 5 (high)
64.7
10
Gemini 3.7 Flash (medium)
64.5
11
Kimi K3 (low)
64.5
12
Gemini 3.1 Pro Preview
64.4
13
Gemini 3 Pro Preview (high)
64.3
14
GPT-5.6 Sol (xhigh)
64.2
15
GPT-6.1 Sol (high)
64.1
16
Gemini 3.6 Flash (high)
64.0
17
Qwen3.8 Max (0902)
63.9
18
GPT-5.5 (high)
63.8
19
MiMo-V2.6-Pro
63.5
20
GLM-5.3 (low)
63.5
21
GPT-6 Astra (high)
63.3
22
GPT-5.5 (xhigh)
63.3
23
Gemini 3.5 Flash (medium)
63.2
24
GLM-5.2 (max)
63.2
25
Claude Sonnet 5.5 (xhigh with fallback)
63.1
26
DeepSeek V4.1 Flash (max)
63.1
27
Qwen3.7 Max
63.0
28
GLM-5.3-Flash
63.0
29
Gemini 3 Flash (non-reasoning)
62.9
30
GPT-5.5 Instant (June 2026)
62.9
31
Claude Sonnet 4.6 (non-reasoning, high)
62.8
32
Claude Opus 4.5
62.5
33
GPT-5.6 Terra (xhigh)
62.1
34
GLM-5.1
62.1
35
GPT-5.4 (xhigh)
62.0
36
Claude Sonnet 5 (high)
61.6
37
Qwen3.6 Max Preview
61.5
38
Gemini 3 Flash
61.3
39
GPT-6 Sol (high)
61.3
40
GLM-5
61.2
41
Qwen3.7 Plus
61.0
42
DeepSeek V4 Pro 0813 (max)
60.9
43
Gemini 3.5 Flash-Lite
60.9
44
Hy3
60.7
45
Gemma 4 31BFREE
60.6
46
GPT-5.6 Luna (xhigh)
60.6
47
MiMo-V2.6-Flash
60.5
48
Kimi K2.5
60.4
49
GPT-5.4 mini (xhigh)
60.0
50
Gemini 2.5 Pro
59.9
51
Qwen3.6 Plus
59.6
52
GPT-6 Luna (high)
59.6
53
GPT-4o (Nov)
59.6
54
Qwen3.5 397B A17B
59.5
55
MiniMax-M3
59.4
56
GLM-4.7
59.4
57
GPT-5.1 (high)
59.1
58
Qwen3.8 27B (xhigh)
59.0
59
Gemma 4 26B A4B
59.0
60
DeepSeek V4 Flash 0731 (max)
58.8
61
GPT-5.2 (xhigh)
58.8
62
GPT-5.6 Sol (max)
58.7
63
Qwen3 Max (Preview)
58.6
64
MiMo-V2.5-Pro
58.6
65
GLM 5V Turbo
58.6
66
Gemini 3.1 Flash-Lite
58.5
67
o3
58.4
68
Kimi K2.6
58.2
69
GPT-5 (low)
57.8
70
DeepSeek V3.2 (non-reasoning)
57.6
71
GLM-4.6 (non-reasoning)
57.5
72
Qwen3 Max Thinking (Preview)
57.4
73
Qwen3 235B 2507
57.3
74
DeepSeek V3.2
57.3
75
DeepSeek V3.2 Exp
57.2
76
DeepSeek V3.1 (non-reasoning)
56.8
77
Kimi K2 Thinking
56.8
78
Qwen3.5 122B A10B
56.7
79
DeepSeek V3.1 Terminus
56.6
80
DeepSeek V3.1 Terminus (non-reasoning)
56.5
81
MiniMax-M2.7
56.5
82
Claude Opus 4.8 (max)
56.4
83
Qwen3 VL 235B A22B (指令)
56.3
84
GLM-4.5
56.1
85
Hy3-preview
55.9
86
Qwen3.5 27B
55.8
87
Gemini 2.5 Flash (non-reasoning)
55.2
88
Qwen3 235B (non-reasoning)
55.1
89
GPT-5.4 nano (xhigh)
55.0
90
Qwen3 Next 80B A3B (指令)
54.7
91
DeepSeek R1 0528
54.7
92
DeepSeek V3 0324
54.4
93
Qwen3 VL 235B A22B (推理)
54.3
94
Qwen3.5 35B A3B
54.2
95
Claude Sonnet 4.6 (max)
53.9
96
MiniMax-M2.5
53.9
97
o4-mini (high)
53.9
98
GPT-5 mini (medium)
53.8
99
Qwen3 Coder 480B
53.5
100
MiMo-V2-Flash
53.4
101
MiniMax-M2.1
53.2
102
GPT-4.1 mini
53.0
103
Qwen3 30B (non-reasoning)
52.9
104
Gemini 2.5 Flash-Lite (Sep) (non-reasoning)
52.6
105
GLM-4.6V (non-reasoning)
52.5
106
Qwen3 235B A22B 2507
52.1
107
Gemini 2.5 Flash-Lite (non-reasoning)
52.1
108
Qwen2.5 Max
52.0
109
Claude 3.7 Sonnet (non-reasoning)
51.9
110
GLM-4.5-Air
51.9
111
Qwen3 Next 80B A3B (推理)
51.5
112
Gemma 3 27BFREE
51.1
113
GLM-4.7-FlashFREE
51.0
114
MiniMax M1 80k
50.9
115
o3-mini (high)
50.9
116
DeepSeek V3 (Dec)
50.3
117
gpt-oss-120b (high)
49.6
118
GLM-4.5V (non-reasoning)
49.6
119
o3-mini
49.2
120
Qwen3 32B
49.0
121
GPT-4o (May)
49.0
122
Claude 3.5 Sonnet (Oct)
48.7
123
MiniMax-M2
48.7
124
Gemma 3 12BFREE
48.5
125
GPT-5 nano (high)
48.0
126
o1-preview
47.9
127
GPT-4o (Aug)
47.8
128
Qwen3 30B A3B 2507
46.8
129
GPT-4 Turbo
46.6
130
GPT-4.1 nano
46.4
131
Claude 3 Opus
46.3
132
gpt-oss-20b (high)
45.8
133
GPT-4o mini
45.8
134
GPT-4.1
45.3
135
Gemma 3 4BFREE
44.3
136
Qwen2.5 72B
44.2
137
GPT-4
41.3
138
Claude 3 Haiku
39.7
139
GPT-3.5 Turbo
33.4
140
QwQ-32B
27.9