跳到主要内容

每日同步 Arena 公开数据

AI 模型排行榜

从匿名对战投票与真实任务反馈,了解 AI 模型的表现。覆盖文本、编程、视觉、图像和视频 12 个榜单。

最近更新
2026-10-12
上榜模型
712
榜单
12
怎么排的
总览/Agent

Agent 榜

真实任务中的执行、工具使用与用户反馈

数据截至 52 个模型
近 8 期选择模型
1
Anthropic新上榜
17.89%±4.24%
5,686$4 / $201M—
2
OpenAI↑3
13.81%±2.34%
11,376$2 / $101.1M2026-09-24:—;2026-09-25:—;2026-09-27:—;2026-09-28:—;2026-09-30:—;2026-10-02:11.23%;2026-10-08:11.72%;2026-10-12:13.81%
3
OpenAI↑3
12.60%±2.60%
6,148$2 / $101.1M2026-09-24:—;2026-09-25:7.68%;2026-09-27:8.18%;2026-09-28:8.80%;2026-09-30:10.65%;2026-10-02:9.71%;2026-10-08:9.92%;2026-10-12:12.60%
4
Google↑3
12.37%±1.94%
24,314——2026-09-24:—;2026-09-25:—;2026-09-27:—;2026-09-28:—;2026-09-30:7.92%;2026-10-02:7.57%;2026-10-08:9.26%;2026-10-12:12.37%
5
OpenAI↓3
12.34%±2.47%
12,483$10 / $501.1M2026-09-24:11.08%;2026-09-25:10.85%;2026-09-27:10.31%;2026-09-28:10.36%;2026-09-30:12.18%;2026-10-02:12.27%;2026-10-08:13.09%;2026-10-12:12.34%
6
Anthropic↓3
11.74%±2.10%
16,099$10 / $501M2026-09-24:13.44%;2026-09-25:13.80%;2026-09-27:13.84%;2026-09-28:14.06%;2026-09-30:14.55%;2026-10-02:14.31%;2026-10-08:12.66%;2026-10-12:11.74%
7
Anthropic↓3
10.75%±3.29%
7,352$2 / $101M2026-09-24:—;2026-09-25:—;2026-09-27:—;2026-09-28:—;2026-09-30:—;2026-10-02:12.52%;2026-10-08:11.95%;2026-10-12:10.75%
8
Anthropic
9.27%±1.41%
42,511$10 / $501M2026-09-24:8.37%;2026-09-25:8.28%;2026-09-27:8.05%;2026-09-28:7.99%;2026-09-30:8.37%;2026-10-02:8.21%;2026-10-08:8.94%;2026-10-12:9.27%
9
Anthropic
8.21%±1.59%
23,430$5 / $251M2026-09-24:9.54%;2026-09-25:9.51%;2026-09-27:9.58%;2026-09-28:9.54%;2026-09-30:8.55%;2026-10-02:7.92%;2026-10-08:8.12%;2026-10-12:8.21%
10
Anthropic↑1
7.76%±1.48%
40,304$5 / $251M2026-09-24:7.27%;2026-09-25:7.31%;2026-09-27:7.02%;2026-09-28:6.87%;2026-09-30:6.92%;2026-10-02:6.64%;2026-10-08:7.13%;2026-10-12:7.76%
11
OpenAI↑1
7.61%±1.41%
36,446$4 / $201.1M2026-09-24:6.20%;2026-09-25:6.16%;2026-09-27:6.21%;2026-09-28:6.26%;2026-09-30:7.05%;2026-10-02:6.45%;2026-10-08:6.23%;2026-10-12:7.61%
12
Google↑10
5.18%±1.07%
37,393$0.75 / $3.751M2026-09-24:4.01%;2026-09-25:4.11%;2026-09-27:4.07%;2026-09-28:3.75%;2026-09-30:2.96%;2026-10-02:2.97%;2026-10-08:2.75%;2026-10-12:5.18%
13
OpenAI↑1
4.71%±1.19%
54,389$5 / $301.1M2026-09-24:4.44%;2026-09-25:4.36%;2026-09-27:4.30%;2026-09-28:4.24%;2026-09-30:4.12%;2026-10-02:4.15%;2026-10-08:4.38%;2026-10-12:4.71%
14
Anthropic↓1
4.35%±1.79%
32,993$2 / $101M2026-09-24:4.84%;2026-09-25:4.79%;2026-09-27:4.82%;2026-09-28:4.80%;2026-09-30:4.81%;2026-10-02:4.40%;2026-10-08:4.91%;2026-10-12:4.35%
15
SpaceXAI↑6
3.59%±1.76%
14,522$2 / $6500K2026-09-24:3.96%;2026-09-25:3.79%;2026-09-27:4.38%;2026-09-28:4.02%;2026-09-30:3.64%;2026-10-02:4.04%;2026-10-08:2.83%;2026-10-12:3.59%
16
DeepSeek开源
3.34%±0.41%
210,741$0.15 / $0.601M2026-09-24:4.06%;2026-09-25:4.10%;2026-09-27:3.96%;2026-09-28:3.86%;2026-09-30:4.01%;2026-10-02:4.02%;2026-10-08:3.80%;2026-10-12:3.34%
17
Meta↓2
3.26%±0.83%
73,262$1.25 / $4.251M2026-09-24:4.03%;2026-09-25:3.90%;2026-09-27:3.96%;2026-09-28:3.83%;2026-09-30:4.19%;2026-10-02:4.00%;2026-10-08:4.04%;2026-10-12:3.26%
18
Xiaomi开源↑1
3.08%±1.34%
20,697$0.43 / $0.871M2026-09-24:—;2026-09-25:—;2026-09-27:—;2026-09-28:—;2026-09-30:—;2026-10-02:—;2026-10-08:3.52%;2026-10-12:3.08%
19
Z.ai开源↑1
3.08%±0.82%
86,621$1.40 / $4.401M2026-09-24:3.98%;2026-09-25:3.99%;2026-09-27:3.91%;2026-09-28:3.84%;2026-09-30:3.64%;2026-10-02:3.48%;2026-10-08:3.19%;2026-10-12:3.08%
20
Moonshot开源↓3
2.99%±0.70%
134,700$3 / $151M2026-09-24:4.56%;2026-09-25:4.55%;2026-09-27:4.24%;2026-09-28:4.03%;2026-09-30:3.98%;2026-10-02:4.18%;2026-10-08:3.78%;2026-10-12:2.99%
21
DeepSeek开源↑5
2.97%±1.82%
7,163——2026-09-24:3.29%;2026-09-25:3.28%;2026-09-27:3.19%;2026-09-28:1.74%;2026-09-30:1.29%;2026-10-02:1.19%;2026-10-08:1.34%;2026-10-12:2.97%
22
OpenAI↑3
2.76%±0.54%
144,508$0.10 / $0.501.1M2026-09-24:—;2026-09-25:—;2026-09-27:—;2026-09-28:1.59%;2026-09-30:1.68%;2026-10-02:1.35%;2026-10-08:1.93%;2026-10-12:2.76%
23
OpenAI↑5
2.67%±1.03%
81,887$2.50 / $151.1M2026-09-24:0.49%;2026-09-25:0.43%;2026-09-27:0.37%;2026-09-28:0.31%;2026-09-30:0.91%;2026-10-02:1.05%;2026-10-08:1.19%;2026-10-12:2.67%
24
Tencent开源↓6
2.46%±0.71%
79,926$0.83 / $2.501M2026-09-24:4.40%;2026-09-25:4.37%;2026-09-27:4.25%;2026-09-28:4.10%;2026-09-30:4.24%;2026-10-02:3.95%;2026-10-08:3.61%;2026-10-12:2.46%
25
Xiaomi开源↑7
2.32%±1.60%
18,238$0.14 / $0.281M2026-09-24:—;2026-09-25:—;2026-09-27:—;2026-09-28:—;2026-09-30:—;2026-10-02:-0.57%;2026-10-08:0.35%;2026-10-12:2.32%

已显示 25 / 52 个模型。细线为 95% 置信区间;较上一期与近 8 期仅比较相同场景和口径,不跨计分调整日。

数据来源:Arena(arena.ai),CC BY 4.0,经万米翻译整理

价格参考 models.dev,单位为美元 / 百万 token,按输入 / 输出展示。厂商图标来自 Lobe Icons。

排名方法

真实偏好,公开数据

大多数榜单来自用户对匿名模型回答的比较。Agent 榜以真实任务中的反馈衡量净提升,用百分比表示。名次、分数和误差均取自 Arena 发布的数据集。

如何阅读分数

条形末端的细线是 95% 置信区间。区间重叠意味着差异仍有不确定性;名次区间按同一期完整分类的置信区间推导,不改变数据集名次。

不同口径,分别比较

文本榜可选择风格校正、原始分与事实性。风格校正减少回答长度和排版的影响。不同子集、场景与推理档位分别展示,不合并排名。

发布与走势

每个榜单标注自己的发布日期。“较上一期”和近 8 期按相同子集、相同分类的发布计算,不跨越 2025-05-16 的计分调整日;缺失期不插值。上榜模型数按当前各榜的模型原名与厂商去重。