M
ModelSage
排行榜/Agent 智能体

AI Agent能力排行榜

基于AA Agentic Index + Arena Agent评分，找到自主执行任务最强的大模型。

分数怎么算：分数 = 该模型的 AA Agentic Index ÷ 当前最高 Index × 100，榜首正好 100 分。Arena 的真人 Agent 会话榜在「Arena 人类盲测 / Agent 智能体」单独看，两者量纲不同不混算。

数据更新于 2026-10-14，每天北京时间凌晨自动刷新

综合能力代码能力Agent 智能体Arena 人类偏好视觉/多模态搜索能力文档理解响应速度性价比
RankModelScore
1
Claude Opus 5.5 (max with fallback)
100.0
2
Claude Sonnet 5.5 (max with fallback)
97.8
3
Claude Opus 5.5 (xhigh with fallback)
97.7
4
Claude Fable 5.1 (max with fallback)
91.6
5
Claude Fable 5.1 (xhigh with fallback)
89.9
6
Claude Sonnet 5.5 (xhigh with fallback)
89.6
7
Claude Opus 5 (max)
89.0
8
Grok 4.7 (xhigh)
88.2
9
Grok 4.7 (high)
87.9
10
Claude Opus 5.5 (high with fallback)
87.6
11
Claude Opus 5 (xhigh)
86.7
12
MiMo-V2.6-Pro
86.0
13
Qwen3.8 Max (0902)
84.9
14
GLM-5.3 (max)
83.6
15
Grok 4.6 (xhigh)
83.3
16
GLM-5.3-Flash
83.2
17
Claude Fable 5.1 (high with fallback)
82.3
18
Qwen3.8-Flash-Next
82.0
19
Gemini 4 Argon (high)
81.6
20
Claude Haiku 5.5 (max)
81.3
21
Grok 4.6 (high)
81.2
22
Grok 4.6 (medium)
80.8
23
Qwen3.8 2.4T A95B
80.5
24
Claude Fable 5 (with fallback)
80.5
25
GPT-5.6 Sol (max)
80.4
26
MiMo-V2.6-Flash
80.2
27
Qwen3.8 Max
80.1
28
DeepSeek V4.1 Flash (max)
79.8
29
Claude Opus 5 (high)
79.4
30
GPT-6.1 Sol (max)
79.2
31
Grok 4.7 (low)
78.9
32
Claude Opus 5.5 (medium with fallback)
78.5
33
GPT-6 Astra (max)
77.6
34
GPT-5.6 Sol (xhigh)
77.4
35
GPT-6 Astra (xhigh)
76.3
36
Claude Sonnet 5.5 (high with fallback)
76.0
37
DeepSeek V4 Flash Vision (max)
75.7
38
Claude Fable 5.1 (medium with fallback)
75.6
39
GPT-6.1 Sol (xhigh)
75.1
40
Kimi K3 (max)
74.9
41
GPT-6 Astra (high)
74.0
42
Claude Haiku 5.5 (xhigh)
73.0
43
GPT-6.1 Sol (high)
72.9
44
GPT-6 Sol (max)
72.8
45
GPT-5.6 Sol (high)
72.5
46
Claude Opus 5 (medium)
71.4
47
GPT-6 Astra (medium)
71.1
48
Claude Fable 5.1 (low with fallback)
69.8
49
Claude Sonnet 5 (max)
69.4
50
GPT-5.6 Luna (max)
69.1
51
GPT-6 Sol (xhigh)
68.8
52
Claude Opus 4.8 (max)
68.6
53
GPT-5.6 Terra (max)
68.5
54
DeepSeek V4 Pro 0813 (max)
68.3
55
GPT-6.1 Sol (medium)
68.3
56
GPT-5.6 Terra (xhigh)
68.1
57
Gemini 3.8 Flash (high)
67.3
58
DeepSeek V4 Flash 0731 (max)
67.2
59
GPT-6 Luna (max)
66.9
60
Gemini 3.8 Flash (medium)
66.3
61
Mistral Large 4 Preview
66.3
62
GPT-5.6 Sol (medium)
66.2
63
Qwen3.8 27B (xhigh)
66.0
64
Claude Haiku 5.5 (high)
65.9
65
Grok 4.6 (low)
65.2
66
GPT-5.6 Luna (xhigh)
64.5
67
GPT-6 Sol (high)
64.1
68
Gemini 3.7 Flash (high)
64.0
69
GPT-6 Astra (low)
63.7
70
Grok 4.5 (high)
63.5
71
Qwen3.8 27B (medium)
63.2
72
GPT-5.6 Terra (high)
62.8
73
GLM-5.2 (max)
62.4
74
GPT-6 Luna (xhigh)
61.8
75
Qwen3.8 27B (low)
61.7
76
Claude Sonnet 5 (xhigh)
61.4
77
Gemini 3.7 Flash (medium)
61.4
78
Claude Opus 4.7 (max)
61.0
79
GPT-5.5 (xhigh)
60.8
80
GPT-6 Sol (medium)
60.4
81
DeepSeek V4.1 Flash (non-reasoning)
60.1
82
GPT-6 Luna (high)
60.0
83
GPT-5.6 Luna (high)
59.2
84
GPT-5.5 (high)
59.0
85
Gemini 3.7 Flash (low)
58.8
86
Claude Sonnet 5.5 (medium with fallback)
58.7
87
GLM-5.3 (low)
58.7
88
GPT-6.1 Sol (low)
58.2
89
Gemini 3.8 Flash (low)
57.5
90
GPT-5.6 Sol (low)
57.3
91
Claude Opus 5 (low)
57.1
92
Gemini 3.6 Flash (high)
55.7
93
Claude Haiku 5.5 (medium)
55.0
94
GPT-5.6 Terra (medium)
54.3
95
GPT-6 Luna (medium)
54.1
96
Claude Sonnet 5 (high)
53.5
97
GPT-6 Sol (non-reasoning)
53.1
98
GLM-5.2 (non-reasoning)
53.0
99
GPT-5.4 (xhigh)
52.9
100
MiniMax-M3
52.6
101
GPT-5.6 Sol (non-reasoning)
52.6
102
Claude Opus 5.5 (low with fallback)
51.9
103
Claude Sonnet 4.6 (max)
51.8
104
GPT-5.5 (medium)
51.7
105
Claude Sonnet 5 (non-reasoning)
50.2
106
Gemini 3.5 Flash (high)
49.6
107
GPT-6 Sol (low)
49.4
108
Claude Sonnet 5.5 (low with fallback)
47.5
109
Claude Sonnet 5 (medium)
46.1
110
DeepSeek V4 Pro (max)
46.0
111
DeepSeek V4 Pro (high)
45.6
112
GPT-5.6 Luna (medium)
44.0
113
Qwen3.7 Max
43.9
114
Kimi K3 (low)
43.8
115
Claude Haiku 5.5 (low)
43.5
116
MiMo-V2.5-Pro
43.4
117
GLM-5.1
43.0
118
GPT-5.6 Terra (low)
42.4
119
GPT-5.6 Terra (non-reasoning)
41.6
120
Qwen3.8 27B (non-reasoning)
39.7
121
Hy3
39.3
122
Claude Sonnet 5 (low)
39.2
123
Grok Build 0.1 0616
39.2
124
DeepSeek V4 Pro 0813 (non-reasoning)
38.6
125
GPT-5.5 (low)
37.6
126
GPT-6 Luna (non-reasoning)
37.6
127
Kimi K2.6
37.1
128
Kimi K2.7 Code
37.0
129
DeepSeek V4 Flash (max)
36.9
130
GPT-6 Luna (low)
36.7
131
GLM-4.7
35.2
132
GPT-5.4 mini (xhigh)
35.1
133
MiniMax-M2.7
34.9
134
DeepSeek V4 Flash (high)
34.5
135
MiMo-V2.5
33.9
136
GPT-5.6 Luna (low)
33.8
137
Qwen3.6 Plus
33.4
138
Qwen3.6 27B
33.1
139
Gemini 3.5 Flash-Lite
32.9
140
GPT-5.5 (non-reasoning)
31.8
141
Qwen3.6 27B (non-reasoning)
31.4
142
GPT-5.4 nano (xhigh)
30.2
143
Grok 4.3 (non-reasoning)
29.8
144
Grok 4.3
29.4
145
GPT-5.6 Luna (non-reasoning)
28.0
146
GPT-5 (high)
27.9
147
Claude 4.5 Sonnet
27.3
148
Qwen3.6 35B A3B
26.0
149
Qwen3.6 35B A3B (non-reasoning)
22.5
150
Kimi K2.5
22.0
151
GPT-5.1 (high)
21.0
152
Qwen3.5 122B A10B
20.0
153
Qwen3.5 397B A17B
18.4
154
Gemini 3.1 Pro Preview
18.3
155
Qwen3.7 Plus
16.5
156
GPT-5 mini (high)
16.4
157
Mistral Medium 3.5
16.0
158
GLM-4.6
15.7
159
Claude 4.5 Haiku
13.8
160
Qwen3.5 122B A10B (non-reasoning)
12.1
161
DeepSeek V3.1 Terminus
12.0
162
DeepSeek V3.2
10.7
163
Gemma 4 31BFREE
5.3
164
Qwen3.5 35B A3B (non-reasoning)
4.4
165
gpt-oss-120b (high)
4.4
166
GPT-5.4 mini (non-reasoning)
3.0
167
Gemma 4 26B A4B
1.2
168
Qwen3 235B A22B 2507
0.0
169
Qwen3.5 9B
0.0
170
Qwen3 Next 80B A3B (推理)
0.0
171
Qwen3 30B A3B 2507
0.0
172
Qwen3 Coder Next
0.0
173
Qwen3 32B
0.0
174
Qwen3 14B
0.0
175
Qwen3 8B
0.0
176
Qwen3.5 2B
0.0
177
Qwen3.5 2B (non-reasoning)
0.0
178
Qwen3.5 0.8B
0.0
179
Qwen3.5 0.8B (non-reasoning)
0.0
180
DeepSeek R1 (Jan)
0.0
181
DeepSeek V3 0324
0.0
182
DeepSeek V3 (Dec)
0.0
183
Gemini 2.5 Pro
0.0
184
Gemini 3.1 Flash-Lite
0.0
185
Gemma 4 12B
0.0
186
Gemma 4 31B (non-reasoning)
0.0
187
Gemma 4 E4B
0.0
188
Gemma 3 27BFREE
0.0
189
Gemma 3 12BFREE
0.0
190
Mistral Small 4
0.0
191
Mistral Large 3
0.0
192
Magistral Small 1.2
0.0
193
Mistral Small 3.2
0.0
194
Mistral Small 3.1
0.0
195
Ministral 3 14B
0.0
196
Ministral 3 8B
0.0
197
Ministral 3 3B
0.0
198
GPT-5.5 Instant (June 2026)
0.0
199
o3-mini (high)
0.0
200
gpt-oss-120b (low)
0.0
201
GPT-4.1 mini
0.0
202
gpt-oss-20b (high)
0.0
203
GPT-4o (Nov)
0.0
204
GPT-4.1 nano
0.0
205
GPT-4
0.0
206
GPT-4o mini
0.0