大模型排行榜
基准跑分与人类盲测两套口径对比 · Artificial Analysis 5 张榜 / 386 个模型 · Arena AI 29 个榜单 / 2,966 条记录
图像与视频模型能力排行 · 163 个模型 · 按张 / 按秒计价 →基准跑分
Artificial Analysis 客观评测 · 标准化测试集 · 可复现 · 更新于 2026-10-14
Artificial Analysis
综合能力Intelligence Index · 384 个模型
完整榜单 →Model
分数输入输出
🥇
Claude Opus 5.5 (max with fallback)Anthropic
100.0¥26.8¥134
🥈
Claude Sonnet 5.5 (max with fallback)Anthropic
97.2¥13.4¥66.9
🥉
Claude Opus 5.5 (xhigh with fallback)Anthropic
97.2¥26.8¥134
4
Claude Opus 5.5 (high with fallback)Anthropic
93.0¥26.8¥134
5
Claude Fable 5.1 (max with fallback)Anthropic
92.6¥66.9¥335
6
Claude Fable 5.1 (xhigh with fallback)Anthropic
92.3¥66.9¥335
7
GPT-6 Astra (max)OpenAI
91.4¥66.9¥335
8
Gemini 4 Argon (high)Google
91.2¥13.4¥66.9
9
GPT-6 Astra (xhigh)OpenAI
90.9¥66.9¥335
10
Claude Sonnet 5.5 (xhigh with fallback)Anthropic
90.1¥13.4¥66.9
11
GPT-6.1 Sol (max)OpenAI
90.0¥13.4¥66.9
12
Claude Opus 5.5 (medium with fallback)Anthropic
88.9¥26.8¥134
13
Claude Fable 5.1 (high with fallback)Anthropic
88.8¥66.9¥335
14
GPT-6.1 Sol (xhigh)OpenAI
88.6¥13.4¥66.9
15
GPT-6 Astra (high)OpenAI
88.4¥66.9¥335
16
Claude Opus 5 (max)Anthropic
88.1¥33.5¥167
17
GPT-6.1 Sol (high)OpenAI
87.2¥13.4¥66.9
18
Claude Opus 5 (xhigh)Anthropic
86.2¥33.5¥167
19
Claude Fable 5 (with fallback)Anthropic
86.1¥66.9¥335
20
GPT-6 Astra (medium)OpenAI
86.0¥66.9¥335
21
Claude Fable 5.1 (medium with fallback)Anthropic
84.9¥66.9¥335
22
Claude Opus 5 (high)Anthropic
83.5¥33.5¥167
23
GPT-6.1 Sol (medium)OpenAI
82.9¥13.4¥66.9
24
GPT-6 Sol (max)OpenAI
82.7¥13.4¥66.9
25
GPT-5.6 Sol (max)OpenAI
81.5¥26.8¥134
26
Claude Fable 5.1 (low with fallback)Anthropic
81.2¥66.9¥335
27
Claude Sonnet 5.5 (high with fallback)Anthropic
81.1¥13.4¥66.9
代码能力Coding Index · 308 个模型
完整榜单 →Model
分数输入输出
🥇
Claude Fable 5.1 (max with fallback)Anthropic
100.0¥66.9¥335
🥈
Claude Fable 5.1 (xhigh with fallback)Anthropic
98.8¥66.9¥335
🥉
Claude Fable 5.1 (high with fallback)Anthropic
97.0¥66.9¥335
4
GPT-5.6 Sol (xhigh)OpenAI
96.1¥26.8¥134
5
GPT-6 Astra (high)OpenAI
95.8¥66.9¥335
6
Claude Opus 5 (max)Anthropic
95.5¥33.5¥167
7
GPT-6 Astra (xhigh)OpenAI
95.2¥66.9¥335
8
GPT-6 Astra (max)OpenAI
94.9¥66.9¥335
9
GPT-6 Astra (medium)OpenAI
94.9¥66.9¥335
10
GPT-5.6 Sol (max)OpenAI
94.9¥26.8¥134
11
Grok 4.6 (high)xAI
94.9¥13.4¥40.2
12
Claude Fable 5.1 (medium with fallback)Anthropic
94.6¥66.9¥335
13
GPT-5.6 Sol (high)OpenAI
94.5¥26.8¥134
14
Gemini 3.8 Flash (high)Google
94.4¥5.0¥25.1
15
Claude Opus 5 (xhigh)Anthropic
94.3¥33.5¥167
16
GPT-5.6 Terra (max)OpenAI
94.0¥13.4¥80.3
17
Claude Opus 5 (high)Anthropic
93.9¥33.5¥167
18
GPT-6 Astra (low)OpenAI
93.7¥66.9¥335
19
Claude Fable 5 (with fallback)Anthropic
93.6¥66.9¥335
20
Qwen3.8 Max (0902)阿里百炼
93.5¥12.0¥36.0
21
GPT-5.6 Sol (medium)OpenAI
93.5¥26.8¥134
22
Kimi K3 (max)Moonshot/Kimi
93.4¥20.0¥100
23
Grok 4.6 (xhigh)xAI
93.3¥13.4¥40.2
24
Gemini 3.7 Flash (high)Google
93.1¥5.0¥25.1
25
Claude Fable 5.1 (low with fallback)Anthropic
92.3¥66.9¥335
26
GLM-5.3 (max)智谱AI
92.2¥8.0¥28.0
27
GPT-5.5 (xhigh)OpenAI
91.3¥33.5¥201
Agent 智能体Agentic Index · 206 个模型
完整榜单 →Model
分数输入输出
🥇
Claude Opus 5.5 (max with fallback)Anthropic
100.0¥26.8¥134
🥈
Claude Sonnet 5.5 (max with fallback)Anthropic
97.8¥13.4¥66.9
🥉
Claude Opus 5.5 (xhigh with fallback)Anthropic
97.7¥26.8¥134
4
Claude Fable 5.1 (max with fallback)Anthropic
91.6¥66.9¥335
5
Claude Fable 5.1 (xhigh with fallback)Anthropic
89.9¥66.9¥335
6
Claude Sonnet 5.5 (xhigh with fallback)Anthropic
89.6¥13.4¥66.9
7
Claude Opus 5 (max)Anthropic
89.0¥33.5¥167
8
Grok 4.7 (xhigh)xAI
88.2¥13.4¥40.2
9
Grok 4.7 (high)xAI
87.9¥13.4¥40.2
10
Claude Opus 5.5 (high with fallback)Anthropic
87.6¥26.8¥134
11
Claude Opus 5 (xhigh)Anthropic
86.7¥33.5¥167
12
MiMo-V2.6-Pro小米
86.0¥2.9¥5.8
13
Qwen3.8 Max (0902)阿里百炼
84.9¥12.0¥36.0
14
GLM-5.3 (max)智谱AI
83.6¥8.0¥28.0
15
Grok 4.6 (xhigh)xAI
83.3¥13.4¥40.2
16
GLM-5.3-Flash智谱AI
83.2¥0.80¥2.8
17
Claude Fable 5.1 (high with fallback)Anthropic
82.3¥66.9¥335
18
Qwen3.8-Flash-Next阿里百炼
82.0¥1.0¥3.1
19
Gemini 4 Argon (high)Google
81.6¥13.4¥66.9
20
Claude Haiku 5.5 (max)Anthropic
81.3¥0.67¥3.3
21
Grok 4.6 (high)xAI
81.2¥13.4¥40.2
22
Grok 4.6 (medium)xAI
80.8¥13.4¥40.2
23
Qwen3.8 2.4T A95B阿里百炼
80.5¥12.0¥36.0
24
Claude Fable 5 (with fallback)Anthropic
80.5¥66.9¥335
25
GPT-5.6 Sol (max)OpenAI
80.4¥26.8¥134
26
MiMo-V2.6-Flash小米
80.2¥0.94¥1.9
27
Qwen3.8 Max阿里百炼
80.1¥12.0¥36.0
响应速度Output Speed 百分位 · 301 个模型
完整榜单 →Model
分数输入输出
🥇
Gemini 3.5 Flash-LiteGoogle
100.0¥2.0¥16.7
🥈
Gemini 3.1 Flash-LiteGoogle
99.7¥1.7¥10.0
🥉
Gemini 3.7 Flash (low)Google
99.3¥5.0¥25.1
4
Gemini 2.5 Flash-LiteGoogle
99.0¥0.67¥2.7
5
Gemini 3.7 Flash (high)Google
98.7¥5.0¥25.1
6
Gemini 3.7 Flash (medium)Google
98.3¥5.0¥25.1
7
gpt-oss-20b (low)OpenAI
98.0¥0.47¥1.4
8
Gemini 2.5 Flash-Lite (non-reasoning)Google
97.7¥0.67¥2.7
9
gpt-oss-120b (low)OpenAI
97.3¥1.0¥3.3
10
Claude Haiku 5.5 (max)Anthropic
97.0¥0.67¥3.3
11
GPT-5.4 mini (xhigh)OpenAI
96.7¥5.0¥30.1
12
gpt-oss-120b (high)OpenAI
96.3¥1.0¥3.9
13
Qwen3.5 Omni Flash阿里百炼
96.0¥2.2¥13.3
14
GPT-4.1 nanoOpenAI
95.7¥0.67¥2.7
15
gpt-oss-20b (high)OpenAI
95.3¥0.47¥1.2
16
Ministral 3 3BMistral
95.0¥0.67¥0.67
17
Gemini 2.5 FlashGoogle
94.7¥2.0¥16.7
18
Gemini 2.5 Flash (non-reasoning)Google
94.4¥2.0¥16.7
19
Gemini 3 Flash (non-reasoning)Google
94.0¥3.3¥20.1
20
GPT-5.4 mini (medium)OpenAI
93.7¥5.0¥30.1
21
GPT-5 nano (high)OpenAI
93.4¥0.33¥2.7
22
o3-mini (high)OpenAI
93.0¥7.4¥29.5
23
Qwen3.7 Max阿里百炼
92.7¥12.0¥36.0
24
GPT-5.4 mini (non-reasoning)OpenAI
92.4¥5.0¥30.1
25
Gemini 3.5 Flash (medium)Google
92.0¥10.0¥60.2
26
Gemini 3.6 Flash (high)Google
91.7¥5.0¥25.1
27
DeepSeek V4.1 Flash (non-reasoning)DeepSeek
91.4¥2.0¥8.0
性价比能力 ÷ 混合价格 · 能力分 ≥ 65 · 71 个模型
完整榜单 →Model
分数输入输出
🥇
Claude Haiku 5.5 (max)Anthropic
100.0¥0.67¥3.3
🥈
MiMo-V2.6-Flash小米
100.0¥0.94¥1.9
🥉
GLM-5.3-Flash智谱AI
99.0¥0.80¥2.8
4
Claude Haiku 5.5 (xhigh)Anthropic
95.0¥0.67¥3.3
5
GPT-6 Luna (max)OpenAI
88.0¥0.67¥3.3
6
Claude Haiku 5.5 (high)Anthropic
87.0¥0.67¥3.3
7
Qwen3.8-Flash-Next阿里百炼
80.0¥1.0¥3.1
8
MiMo-V2.6-Pro小米
39.0¥2.9¥5.8
9
DeepSeek V4.1 Flash (max)DeepSeek
35.0¥2.0¥8.0
10
Gemini 3.8 Flash (high)Google
13.0¥5.0¥25.1
11
Gemini 3.8 Flash (medium)Google
12.0¥5.0¥25.1
12
Gemini 3.7 Flash (medium)Google
12.0¥5.0¥25.1
13
Gemini 3.7 Flash (high)Google
12.0¥5.0¥25.1
14
GLM-5.3 (max)智谱AI
11.0¥8.0¥28.0
15
Qwen3.8 Max (0902)阿里百炼
8.0¥12.0¥36.0
16
Grok 4.7 (xhigh)xAI
7.0¥13.4¥40.2
17
Grok 4.7 (high)xAI
7.0¥13.4¥40.2
18
Qwen3.8 Max阿里百炼
7.0¥12.0¥36.0
19
Qwen3.8 2.4T A95B阿里百炼
7.0¥12.0¥36.0
20
Grok 4.6 (high)xAI
7.0¥13.4¥40.2
21
Grok 4.6 (xhigh)xAI
7.0¥13.4¥40.2
22
Grok 4.6 (medium)xAI
7.0¥13.4¥40.2
23
Grok 4.7 (low)xAI
6.0¥13.4¥40.2
24
Claude Sonnet 5.5 (max with fallback)Anthropic
6.0¥13.4¥66.9
25
Gemini 4 Argon (high)Google
6.0¥13.4¥66.9
26
Claude Sonnet 5.5 (xhigh with fallback)Anthropic
6.0¥13.4¥66.9
27
GPT-6.1 Sol (max)OpenAI
6.0¥13.4¥66.9
人类盲测
Arena AI 真实用户投票 · 匿名对战 · Elo 排名 · 更新于 2026-10-14
Agent 智能体
综合总榜53 个模型 · 5 个信号维度
完整榜单 →Model
综合分数CI会话
🥇
Claude Opus 5.5 (Max)Anthropic
+17.0±3.56.0k
🥈
GPT 6.1 Sol (Max)OpenAI
+15.1±2.112.1k
🥉
Claude Fable 5.1 (Max)Anthropic
+14.3±2.016.2k
4
GPT 6 Astra (Max)OpenAI
+13.7±2.312.8k
5
Gemini 4 Argon (High)Google
+13.6±1.626.9k
6
Claude Sonnet 5.5 (Max)Anthropic
+13.0±3.27.4k
7
GPT 6 Sol (Max)OpenAI
+10.5±2.36.7k
8
Claude Opus 5 (Max)Anthropic
+9.2±1.723.5k
9
Claude Fable 5 (High)Anthropic
+8.5±1.442.4k
10
GPT 5.6 Sol (xHigh)OpenAI
+7.7±1.436.3k
Chat 对话
Text 文本415 个模型
完整榜单 →Model
EloCI投票
🥇
gemini-4-argon-highGoogle
1524.8±8.84.9k
🥈
muse-wonder-maxMeta
1512.7±13.31.9k
🥉
claude-opus-5.5-highAnthropic
1508.1±7.57.7k
4
claude-fable-5-highAnthropic
1503.4±4.341.4k
5
claude-opus-4-6-highAnthropic
1503.1±3.480.8k
6
claude-opus-4-7-highAnthropic
1501±3.865.1k
7
claude-fable-5.1-maxAnthropic
1500.6±6.114.1k
8
claude-opus-4-6Anthropic
1497.7±3.385.4k
9
gemini-3.8-flash-highGoogle
1496.3±4.635.7k
10
muse-spark-1.2 (xHigh)Meta
1494.6±7.47.6k
Search 搜索42 个模型
完整榜单 →Model
EloCI投票
🥇
gpt-5.6-sol-xhighOpenAI
1261.1±6.360.4k
🥈
qwen3.8-maxAlibaba
1261.1±13.21.9k
🥉
claude-opus-4-6-searchAnthropic
1251.4±4.6173.2k
4
gpt-5.5-searchOpenAI
1246.6±4.9126.0k
5
claude-fable-5.1-highAnthropic
1245.8±7.112.8k
6
claude-opus-5-highAnthropic
1241.6±6.913.3k
7
gpt-6-astra-maxOpenAI
1233.6±7.918.5k
8
gpt-6.1-sol-maxOpenAI
1233.1±8.95.6k
9
claude-opus-4-7Anthropic
1230.1±4.9127.5k
10
claude-fable-5-highAnthropic
1229.1±6.573.6k
Vision 视觉165 个模型
完整榜单 →Model
EloCI投票
🥇
claude-fable-5-highAnthropic
1308±7.114.2k
🥈
qwen3.8-maxAlibaba
1300.7±7.410.0k
🥉
claude-opus-4-6-highAnthropic
1299.2±6.522.3k
4
claude-opus-4-7Anthropic
1298.2±6.623.2k
5
claude-opus-4-7-highAnthropic
1297.8±6.522.7k
6
gemini-3.7-flash-highGoogle
1296.9±9.94.5k
7
claude-opus-4-6Anthropic
1294.5±6.426.9k
8
gpt-6.1-sol-maxOpenAI
1294.1±14.51.8k
9
muse-sparkMeta
1293.9±9.15.9k
10
muse-spark-1.2 (xHigh)Meta
1292.9±14.32.0k
Document 文档44 个模型
完整榜单 →Model
EloCI投票
🥇
claude-opus-5-highAnthropic
1516.3±78.8k
🥈
claude-fable-5.1-maxAnthropic
1512.6±15.21.4k
🥉
claude-opus-4-6-highAnthropic
1507.3±6.427.9k
4
claude-opus-4-6Anthropic
1507.1±5.941.3k
5
claude-fable-5-highAnthropic
1496.1±7.98.5k
6
claude-opus-4-7Anthropic
1495±6.422.2k
7
claude-opus-4-7-highAnthropic
1494.7±6.522.0k
8
gpt-5.5OpenAI
1485.7±6.321.3k
9
gpt-5.5-highOpenAI
1484.5±6.420.9k
10
gpt-5.6-sol-xhighOpenAI
1482.7±8.84.8k
Code 代码
Code 编程143 个模型 · 12 个子榜
完整榜单 →Model
EloCI投票
🥇
claude-opus-5.5-maxAnthropic
1811.1±14.62.6k
🥈
gpt-6-astra-maxOpenAI
1780.2±10.55.7k
🥉
claude-sonnet-5.5-xhighAnthropic
1773.7±13.92.9k
4
gpt-6.1-sol-maxOpenAI
1760.1±13.92.6k
5
claude-fable-5.1-maxAnthropic
1742.7±9.17.2k
6
claude-sonnet-5.5-highAnthropic
1713±10.44.6k
7
claude-opus-5-maxAnthropic
1690±6.317.9k
8
gpt-6-sol-maxOpenAI
1685.9±114.0k
9
gemini-4-argon-highGoogle
1678±15.41.8k
10
qwen3.8-max-0902Alibaba
1673.2±7.313.1k
Image-to-WebDev 图生代码64 个模型
完整榜单 →Model
EloCI投票
🥇
claude-opus-5.5-maxAnthropic
1748.5±30.2493
🥈
claude-sonnet-5.5-xhighAnthropic
1740±19.91.1k
🥉
gpt-6-astra-maxOpenAI
1731.4±13.32.7k
4
claude-fable-5.1-maxAnthropic
1719.8±13.72.5k
5
gpt-6.1-sol-maxOpenAI
1703.5±171.4k
6
gpt-6-sol-maxOpenAI
1661.3±13.12.4k
7
claude-opus-5-maxAnthropic
1661.1±10.34.3k
8
muse-spark-1.3-maxMeta
1642±9.84.7k
9
qwen3.8-max-0902Alibaba
1637.9±18.41.2k
10
claude-fable-5-highAnthropic
1622.5±9.25.5k
Image 图像
文生图82 个模型
完整榜单 →Model
EloCI投票
🥇
gpt-image-2.5-sunburstOpenAI
1424.6±6.917.8k
🥈
gpt-image-2.5-flareOpenAI
1397.6±6.816.9k
🥉
gpt-image-2 (medium)OpenAI
1383±493.3k
4
grok-imagine-image-2.0 (canvas)SpaceXAI
1333.5±11.53.1k
5
mai-image-2.6Microsoft AI
1332.1±5.425.1k
6
gemini-nano-banana-2.1Google
1327.8±8.85.3k
7
reve-2.1Reve
1301.7±7.58.2k
8
grok-imagine-image-2.0 (20260801)SpaceXAI
1294.2±6.314.5k
9
muse-imageMeta
1274±5.142.3k
10
reve-2.0Reve
1269.2±6.215.8k
图像编辑59 个模型 · 2 个子榜
完整榜单 →Model
EloCI投票
🥇
gpt-image-2.5-sunburstOpenAI
1523.6±5.459.8k
🥈
gpt-image-2.5-flareOpenAI
1481±5.357.4k
🥉
gpt-image-2 (medium)OpenAI
1461.7±3.2305.8k
4
grok-imagine-image-2.0 (canvas)SpaceXAI
1439.1±8.611.6k
5
mai-image-2.6Microsoft AI
1428.3±551.4k
6
gemini-nano-banana-2.1Google
1428.1±5.813.0k
7
grok-imagine-image-2.0 (20260801)SpaceXAI
1424.5±550.4k
8
muse-imageMeta
1402.6±4.3142.6k
9
mai-image-2.5Microsoft AI
1400.9±3.7190.1k
10
seedream-5.0-proBytedance
1393.6±3.1282.9k
细分场景:多图编辑
Video 视频
文生视频48 个模型
完整榜单 →Model
EloCI投票
🥇
gemini-omni-1.1-flashGoogle
1516.2±15.51.8k
🥈
gemini-omni-flashGoogle
1512.9±8.926.6k
🥉
flux-3-videoBlack Forest Labs
1493.1±17.31.3k
4
grok-imagine-video-1.5-agentSpaceXAI
1492.3±18.31.2k
5
dreamina-seedance-2.0-720pBytedance
1479.2±856.3k
6
wan3.0Alibaba
1476.3±12.72.6k
7
dreamina-seedance-2.5-720pBytedance
1473.9±8.58.0k
8
minimax-h3MiniMax
1460.3±8.611.4k
9
muse-videoMeta
1455.8±14.62.2k
10
happyhorse-1.0Alibaba-ATH
1427.2±12.522.3k
图生视频48 个模型
完整榜单 →Model
EloCI投票
🥇
minimax-h3MiniMax
1495.4±5.257.1k
🥈
gemini-omni-1.1-flashGoogle
1488.5±10.93.7k
🥉
wan3.0Alibaba
1479.8±10.83.4k
4
dreamina-seedance-2.5-720pBytedance
1477.4±7.215.7k
5
dreamina-seedance-2.0-720pBytedance
1474.6±7121.4k
6
gemini-omni-flashGoogle
1464.8±5.687.2k
7
hidream-o1-video-1.0HiDream
1456.5±86.5k
8
grok-imagine-video-1.5-720pSpaceXAI
1456±4.2137.5k
9
flux-3-video-20260811Black Forest Labs
1449.4±537.0k
10
happyhorse-1.0Alibaba-ATH
1441.7±9.970.6k
视频编辑10 个模型
完整榜单 →Model
EloCI投票
🥇
wan3.0Alibaba
1414.1±25.7463
🥈
dreamina-seedance-2.5-720pBytedance
1409.8±26.5429
🥉
minimax-h3MiniMax
1392.2±18.6962
4
gemini-omni-flashGoogle
1366.6±14.62.1k
5
dreamina-seedance-2.0-720pBytedance
1364.6±14.33.9k
6
happyhorse-1.0Alibaba-ATH
1306.7±13.63.1k
7
grok-imagine-videoSpaceXAI
1257.8±10.313.7k
8
kling-o3-proKlingAI
1255.3±11.47.5k
9
kling-o1-proKlingAI
1197.3±10.410.5k
10
runway-gen4-alephRunway
1182.2±8.79.8k
Elo = 由真人两两对决胜负累积得到的等级分，仅比较差值：分差 100 分约对应 64% 胜率，差值在个位数以内视为水平相当。CI = 95% 置信区间，投票越多越小；两个模型的 ± 区间重叠，说明名次差别在统计上不成立。Agent 综合分数 = 5 个行为信号加权后相对基线的净提升百分点，+12.0 表示比基线高 12 个百分点，不是百分制评分。
评测方法差异：基准跑分通过标准化测试集（代码、数学、推理等）自动打分，结果客观可复现； 人类盲测由真实用户对匿名模型输出进行两两对比投票，反映主观偏好和实际体验。 Agent 榜不采用投票，而是统计真实 Agent 会话中的行为信号（任务是否被确认完成、用户反馈倾向、纠错效率等）。 三种方法互补，宜结合参考。
数据更新：基准跑分（Artificial Analysis）2026-10-14 · 人类盲测（Arena AI）2026-10-14 · 价格（厂商官网）2026-10-14