v4.3 · 10 项评测

智能指数拆分

解释 AA Intelligence Index v4.3 的 10 项评测、原始指标和权重,按版本查看逐模型成绩并保留 v4.1 历史参考,不跨版本混排或合成新总分。

数据快照:2026-09-16T01:36:22.563Z查看原始来源 ↗
主要指标Evaluation relevance
方法核验日期2026-09-16T01:36:22.563Z

INTELLIGENCE INDEX V4.3

当前 10 项组成评测

01 · Agent · 15%

AA-Briefcase

四个工作场景中的文件交付任务,按任务成功、分析质量和呈现质量进行成对比较。

02 · Agent · 10%

GDPval-AA v2

高经济价值、可交付成果导向的真实知识工作任务。

03 · Agent · 5%

AutomationBench-AA

使用 REST API 完成 SaaS 工作流,触发安全约束违规的任务不得分。

04 · 编码 · 10%

Terminal-Bench v4.0

在终端中执行真实任务,与 v2.1 的任务集和结果分开比较。

05 · 编码 · 10%

SciCode

科研场景中的代码生成、执行和数值计算。

06 · 通用 · 15%

AA-Omniscience

知识准确率与幻觉率分别进入指数;独立 Omniscience Index 不直接作为组成分数。

07 · 通用 · 10%

GDP.pdf

基于长 PDF 回答跨领域问题,保留全通过与任务宏平均通过率的不同口径。

08 · 通用 · 5%

AA-LCR v1.1

跨长上下文定位、保持和综合关键信息。

09 · 科学推理 · 10%

Humanity's Last Exam

覆盖多学科专家级知识与复杂推理。

10 · 科学推理 · 10%

CritPt

要求严谨分析与多步骤推理的科学评测。

按版本展示

逐模型组成评测成绩

展示已核实的来源公开数据,按版本查看并保留旧观测。不同版本不混排;估算保留标记,缺失字段不补算。

528 / 528 个模型

智能指数仍为 v4.3,带 * 为原站估算值;GDPval-AA v2 与 AA-Briefcase 展示实时 Elo,已核实当前裁判协议的值标为 v4.3.1。指数中的 Elo 贡献在模型加入时冻结,不能用实时 Elo 重算指数。旧裁判协议未区分的值保留为历史参考,Elo 排序时置于末尾且不编号。其余分项以百分数展示;幻觉率越低越好,缺失值为 —。

Artificial Analysis Intelligence Index v4.3 完整逐模型分项成绩
#模型机构智能指数GDPval-AA v2 (Elo)AA-Briefcase (Elo)AutomationBench-AATerminal-Bench v4.0SciCodeAA-LCR v1.1AA-Omniscience 准确率AA-Omniscience 幻觉率GDP.pdf 全通过率HLECritPt权重推理发布日期
1Claude Fable 5.1 (max with fallback)Anthropic53.371,723.58
裁判协议 v4.3.1
1,646.09
裁判协议 v4.3.1
59.4%52%63.1%85.3%67.2%72.6%26.2%59.1%29.7%闭源2026-09-01
2Claude Fable 5.1 (xhigh with fallback)Anthropic53.181,745.29
裁判协议 v4.3.1
1,650.01
裁判协议 v4.3.1
57.8%55.1%60.9%83%66.2%70.5%26.2%58.7%31.1%闭源2026-09-01
3GPT-6 Astra (max)OpenAI52.811,580.2
裁判协议 v4.3.1
1,562.01
裁判协议 v4.3.1
68.5%59.1%56.5%80.7%62.6%51.3%31%54.7%31.7%闭源2026-09-03
4GPT-6 Astra (xhigh)OpenAI52.511,555.23
裁判协议 v4.3.1
1,533.73
裁判协议 v4.3.1
67.2%59.6%55.7%80%61.9%48.3%32.2%54.6%31.4%闭源2026-09-03
5Claude Fable 5.1 (high with fallback)Anthropic51.211,649.77
裁判协议 v4.3.1
1,578.19
裁判协议 v4.3.1
55.3%52%58.7%83.7%64.9%68.8%26.8%55.9%30.3%闭源2026-09-01
6GPT-6 Astra (high)OpenAI51.051,529.7
裁判协议 v4.3.1
1,493.87
裁判协议 v4.3.1
66.6%54%55.4%80%61.1%44.8%31%53.1%28.9%闭源2026-09-03
7Claude Opus 5 (max)Anthropic50.71,735.09
裁判协议 v4.3.1
1,633.44
裁判协议 v4.3.1
56.6%49%56.4%79.3%60.9%60.8%21.6%54.9%29.1%闭源2026-07-24
8Claude Fable 5历史Anthropic49.71,631.46
裁判协议 v4.3.1
1,529.62
裁判协议 v4.3.1
54.1%42.4%61%82.3%65.3%63.6%24%55.5%28.6%闭源2026-06-09
9GPT-6 Astra (medium)OpenAI49.671,500.67
裁判协议 v4.3.1
1,450.6
裁判协议 v4.3.1
64.6%49.5%54.2%79.7%60.6%46.5%30.4%52.7%29.1%闭源2026-09-03
10Claude Opus 5 (xhigh)Anthropic49.651,708.14
裁判协议 v4.3.1
1,624.56
裁判协议 v4.3.1
53.2%46.5%55.7%80.3%59.5%59.5%21%54.4%27.7%闭源2026-07-24
11Claude Fable 5.1 (medium with fallback)Anthropic49.061,579.31
裁判协议 v4.3.1
1,531.2
裁判协议 v4.3.1
54.7%44.9%56.4%84.7%63.1%69.1%26.8%53.8%29.1%闭源2026-09-01
12Claude Opus 5 (high)Anthropic48.241,628.94
裁判协议 v4.3.1
1,556.66
裁判协议 v4.3.1
53.6%46%55.4%79%58.9%61.2%19.6%52.8%28.3%闭源2026-07-24
13Muse Spark 1.3 (max)Meta48.171,703.34
裁判协议 v4.3.1
1,589.18
裁判协议 v4.3.1
57.9%33.3%58.8%83%43.6%32.9%26.6%48.7%24.9%闭源2026-09-02
14GPT-5.6 Sol (max)OpenAI47.061,586.78
裁判协议 v4.3.1
1,463.44
裁判协议 v4.3.1
60.1%39.9%57.1%84%59.4%92.2%27.2%49.5%32.3%闭源2026-07-09
15Claude Fable 5.1 (low with fallback)Anthropic47.041,503.72
裁判协议 v4.3.1
1,483.73
裁判协议 v4.3.1
52.2%40.4%56.7%82.3%60.2%65.6%28%48.9%27.7%闭源2026-09-01
16GPT-6 Astra (low)OpenAI45.991,419.29
裁判协议 v4.3.1
1,253.1
裁判协议 v4.3.1
59.1%41.9%54.1%80%59.5%46.9%30.4%49.2%26.3%闭源2026-09-03
17Qwen3.8 Max (0902)Alibaba45.441,664.28
裁判协议 v4.3.1
1,609.69
裁判协议 v4.3.1
56.2%38.9%52.1%80.3%31.7%28.8%22.8%43.1%17.7%闭源2026-09-02
18GPT-6 Astra (Non-reasoning)
历史参考 2026-09-09
OpenAI45.171,542.59
历史参考 · 裁判协议未区分
1,471.32
历史参考 · 裁判协议未区分
61.8%50.5%53.5%70.7%55.6%65.4%26.8%37.1%19.7%闭源2026-09-03
19Muse Spark 1.3 (xhigh)Meta45.161,662.34
裁判协议 v4.3.1
1,483.22
裁判协议 v4.3.1
56.8%16.7%59.7%83%41.5%31.5%24.2%47.5%26%闭源2026-09-02
20Claude Opus 5 (medium)Anthropic45.061,524.87
裁判协议 v4.3.1
1,439.04
裁判协议 v4.3.1
54.3%34.3%51.5%82%57.1%60.7%20%51.3%26.9%闭源2026-07-24
21GLM-5.3 (max)Z AI44.861,633.8
裁判协议 v4.3.1
1,504.38
裁判协议 v4.3.1
62.2%41.9%59%79.7%33.9%29.6%11.2%42.3%19.1%开放2026-08-18
22Grok 4.6 (high)SpaceXAI44.411,642.96
裁判协议 v4.3.1
1,534.13
裁判协议 v4.3.1
66.7%21.2%56.5%80.3%48.2%34.3%17%42.9%17.1%闭源2026-08-12
23Grok 4.6 (xhigh)SpaceXAI44.271,663.07
裁判协议 v4.3.1
1,545.11
裁判协议 v4.3.1
67%17.2%53%81%43%24%17.2%44.1%19.7%闭源2026-08-12
24GPT-5.6 Sol (xhigh)OpenAI44.141,584.88
裁判协议 v4.3.1
1,434.68
裁判协议 v4.3.1
55.3%24.7%57.1%82.3%58.8%91.9%27.6%47.3%28.6%闭源2026-07-09
25Kimi K3 (max)Kimi43.781,547.63
裁判协议 v4.3.1
1,488.1
裁判协议 v4.3.1
58.3%12.6%59.5%88.7%47.6%53.2%22%46.9%23.4%开放2026-07-16
26Grok 4.6 (medium)SpaceXAI43.011,642.56
裁判协议 v4.3.1
1,490.6
裁判协议 v4.3.1
63.2%13.1%55.9%81%41.9%24%17.8%42.1%17.7%闭源2026-08-12
27GPT-5.6 Sol (high)OpenAI42.51,524.01
裁判协议 v4.3.1
1,360.88
裁判协议 v4.3.1
55.3%20.7%57.8%81.7%58.4%91.2%27.8%46%25.7%闭源2026-07-09
28GPT-5.6 Terra (max)OpenAI42.251,476.96
裁判协议 v4.3.1
1,330
裁判协议 v4.3.1
59.6%35.4%55%83%46.8%87.9%24%42.9%30%闭源2026-07-09
29Claude Opus 4.8 (max)历史Anthropic41.991,489.04
裁判协议 v4.3.1
1,315.95
裁判协议 v4.3.1
45.6%21.7%54.4%77.7%48.8%39.3%22.8%48.7%20.9%闭源2026-05-28
30GLM-5.3-FlashZ AI41.911,654.95
裁判协议 v4.3.1
1,448.95
裁判协议 v4.3.1
60.4%32.8%51.6%80%27.5%27.6%15.4%39.9%15.4%开放2026-08-26
31Gemini 3.8 Flash (high)Google41.191,463.81
裁判协议 v4.3.1
1,201.59
裁判协议 v4.3.1
59.9%19.7%56.6%81.3%54.6%55.2%21%47.8%18.3%闭源2026-09-02
32Claude Opus 4.7 (max)历史Anthropic40.69*1,395.78
裁判协议 v4.3.1
1,251.5
裁判协议 v4.3.1
78.7%48.9%42.3%42.3%12%闭源2026-04-16
33Qwen3.8 Max历史Alibaba40.31,630.39
裁判协议 v4.3.1
1,387.89
裁判协议 v4.3.1
49.2%18.7%53.2%78.3%31.9%41.7%20.2%43%20%闭源2026-08-03
34Qwen3.8 2.4T A95BAlibaba40.041,627.82
裁判协议 v4.3.1
1,438.5
裁判协议 v4.3.1
57.2%11.1%54.1%80.3%31.3%39.2%15.6%42.4%20%开放2026-08-12
35Gemini 3.8 Flash (medium)Google401,455.38
裁判协议 v4.3.1
1,176.54
裁判协议 v4.3.1
60.9%19.7%55.1%84%53%51.9%22.8%42.1%12.3%闭源2026-09-02
36Qwen3.8-Flash-NextAlibaba39.911,647.02
裁判协议 v4.3.1
1,586.58
裁判协议 v4.3.1
55.9%25.3%50.6%79.7%24.5%45.3%15.6%38%11.1%开放2026-08-26
37Muse Spark 1.2 (xhigh)历史Meta39.81,523.32
裁判协议 v4.3.1
1,343.5
裁判协议 v4.3.1
40.6%7.1%57.4%79%45.4%33.3%17.4%45.5%17.7%闭源2026-08-05
38Claude Opus 5 (low)Anthropic39.791,371.52
裁判协议 v4.3.1
1,215.73
裁判协议 v4.3.1
51.8%26.3%49.2%81.3%56%62.2%17.2%43.4%23.1%闭源2026-07-24
39Gemini 3.7 Flash (medium)历史Google39.62*1,406.6
裁判协议 v4.3.1
59.8%83%54%65.9%39%9.4%闭源2026-08-13
40DeepSeek V4.1 Flash (max)DeepSeek39.551,632.06
裁判协议 v4.3.1
1,423.72
裁判协议 v4.3.1
68.9%26.8%51.9%84%46.4%96.5%12.8%39.2%14.3%开放2026-09-10
41GPT-5.6 Sol (medium)OpenAI39.481,426.5
裁判协议 v4.3.1
1,236.71
裁判协议 v4.3.1
51.3%14.6%57.4%80.3%57.8%90.8%26.2%42.2%22.9%闭源2026-07-09
42Gemini 3.7 Flash (high)历史Google39.431,433.14
裁判协议 v4.3.1
1,115.67
裁判协议 v4.3.1
62%13.6%57.2%81.7%55.3%64.5%23.6%47.9%14.3%闭源2026-08-13
43Grok 4.5 (high)历史SpaceXAI39.081,429.6
裁判协议 v4.3.1
1,283.38
裁判协议 v4.3.1
57.9%10.6%55%79.3%51.5%54.1%18.8%42.7%15.4%闭源2026-07-08
44GPT-5.4 (xhigh)历史OpenAI38.98*1,306.71
裁判协议 v4.3.1
82%50.8%91.7%43.7%23.4%闭源2026-03-05
45GPT-5.5 (xhigh)历史OpenAI38.631,395.85
裁判协议 v4.3.1
1,135.79
裁判协议 v4.3.1
47.3%14.6%55.8%84.3%58%89%21.2%45.8%27.1%闭源2026-04-23
46Claude Sonnet 5 (max)Anthropic38.361,500.86
裁判协议 v4.3.1
1,355.19
裁判协议 v4.3.1
36.5%14.1%54.3%82%40.1%39.4%13.2%41.3%16.9%闭源2026-06-30
47GPT-5.6 Terra (xhigh)OpenAI38.191,478.6
裁判协议 v4.3.1
1,337.28
裁判协议 v4.3.1
47.1%10.1%52.3%79%45.5%89%24.6%41.9%27.1%闭源2026-07-09
48GPT-5.6 Luna (max)OpenAI37.51,455.6
裁判协议 v4.3.1
1,332.9
裁判协议 v4.3.1
50.2%11.6%53.6%83.7%42.7%92.6%24%39.5%20.6%闭源2026-07-09
49GPT-5.5 (high)历史OpenAI37.291,371.51
裁判协议 v4.3.1
1,088.84
裁判协议 v4.3.1
44.3%9.1%56.1%84.3%57%89.1%22.6%45%25.4%闭源2026-04-23
50Gemini 3.7 Flash (low)历史Google36.95*1,354.06
裁判协议 v4.3.1
55.7%78.7%53.6%67.7%35.1%5.7%闭源2026-08-13
51DeepSeek V4 Pro 0813 (max)DeepSeek36.281,493.31
裁判协议 v4.3.1
1,264.67
裁判协议 v4.3.1
56.7%14.1%51%80.3%49.1%94.8%11.4%41%18%开放2026-08-13
52Agnes 3.0 FlashSapiens AI35.5*1,573.47
裁判协议 v4.3.1
50.7%7.1%51.6%81%25.5%48.4%10%38.5%15.1%闭源2026-09-11
53Grok 4.6 (low)SpaceXAI35.411,461.3
裁判协议 v4.3.1
1,293.13
裁判协议 v4.3.1
53.4%3%49.4%80.7%43.3%30.6%15.8%27.6%5.7%闭源2026-08-12
54Agnes 2.5 Pro BetaSapiens AI35.24*1,375.77
裁判协议 v4.3.1
48.8%83%16.8%32.8%37.5%15.7%闭源2026-08-26
55DeepSeek V4 Flash Vision (max)DeepSeek35.011,577.21
裁判协议 v4.3.1
1,432.36
裁判协议 v4.3.1
47.5%12.1%49.7%81.3%38.6%91.5%12%34.5%10.9%闭源2026-08-21
56GPT-5.6 Luna (xhigh)OpenAI34.771,428.47
裁判协议 v4.3.1
1,268.01
裁判协议 v4.3.1
42.6%3.5%50.5%81.7%42.4%92.5%23.8%37%20.6%闭源2026-07-09
57Claude Sonnet 5 (xhigh)Anthropic34.711,400.77
裁判协议 v4.3.1
1,273.41
裁判协议 v4.3.1
34.5%7.1%54.1%76.7%39%58.6%12.6%39%15.4%闭源2026-06-30
58DeepSeek V4 Flash 0731 (max)历史DeepSeek34.531,441.89
裁判协议 v4.3.1
1,253.08
裁判协议 v4.3.1
54%12.1%50.3%79.7%40.4%91.7%11%38.6%16.6%开放2026-07-31
59GPT-5.6 Terra (high)OpenAI34.491,415.36
裁判协议 v4.3.1
1,201.36
裁判协议 v4.3.1
42%1.5%52.4%77.7%45.5%89.8%20.8%38.5%22.9%闭源2026-07-09
60Gemini 3.6 Flash历史Google34.341,331.37
裁判协议 v4.3.1
952.75
裁判协议 v4.3.1
53%7.1%53.4%80%50%55.6%17.4%40.8%10.6%闭源2026-07-21
61Muse Spark 1.1 (xhigh)历史Meta34.271,294.15
裁判协议 v4.3.1
860.36
裁判协议 v4.3.1
38.8%6.1%58.8%77.7%52%50%14.4%46.2%15.1%闭源2026-07-09
62GPT-5.5 (medium)历史OpenAI34.151,288.18
裁判协议 v4.3.1
1,000
裁判协议 v4.3.1
40.1%5.1%54.5%83%56.8%89.6%20.6%42.4%18.6%闭源2026-04-23
63GLM-5.2 (max)历史Z AI34.011,406.07
裁判协议 v4.3.1
1,231.38
裁判协议 v4.3.1
28.4%1%51.2%78.3%24.3%26.3%10.4%41.1%20.9%开放2026-06-16
64Qwen3.8 27B (xhigh)Alibaba33.91,463.25
裁判协议 v4.3.1
1,397.82
裁判协议 v4.3.1
48.2%5.6%46.6%82%15.6%30.3%16.4%33.9%5.4%开放2026-08-14
65GPT-5.6 Sol (low)OpenAI33.81,354.85
裁判协议 v4.3.1
1,042.06
裁判协议 v4.3.1
41%1%56.4%78%57.2%89.4%21%39.4%14.9%闭源2026-07-09
66Gemini 3.8 Flash (low)Google33.791,358.55
裁判协议 v4.3.1
1,055.59
裁判协议 v4.3.1
36.5%10.1%55%80.7%52.2%64.6%18.6%37.1%4%闭源2026-09-02
67Gemini 3.5 Flash (medium)历史Google33.63*870.71
裁判协议 v4.3.1
74.3%51%61.8%41.3%10.9%闭源2026-05-19
68Motif 3Motif Technologies33.57*1,198.83
裁判协议 v4.3.1
42.2%76%30.1%28.4%37%6.6%开放2026-08-12
69Gemini 3.5 Flash历史Google32.981,259.06
裁判协议 v4.3.1
869.74
裁判协议 v4.3.1
42.1%6.6%53.9%73.3%51.4%62.2%19.8%42.7%13.1%闭源2026-05-19
70GPT-5.3 Codex (xhigh)OpenAI32.5*865.74
裁判协议 v4.3.1
83.3%52.9%89.2%42.5%16.9%闭源2026-02-05
71GPT-5.6 Luna (high)OpenAI32.411,375.25
裁判协议 v4.3.1
1,173.05
裁判协议 v4.3.1
35.6%2.5%51.6%80.3%41.8%92.4%22.2%33.4%16.6%闭源2026-07-09
72Motif 3 (Beta)Motif Technologies32.31*1,182.62
裁判协议 v4.3.1
74.7%21%49.1%40.4%6%闭源2026-07-14
73Claude Sonnet 5 (high)Anthropic32.031,308.45
裁判协议 v4.3.1
1,177.84
裁判协议 v4.3.1
32.1%5.1%54.3%76.7%37.4%65.7%9.4%35.7%15.1%闭源2026-06-30
74Claude Opus 4.6 (max)历史Anthropic31.95*78%47%62.8%39.9%12.6%闭源2026-02-05
75Muse Spark历史Meta31.3*1,076.31
裁判协议 v4.3.1
644.1
裁判协议 v4.3.1
78%49.6%84.2%40.7%11.3%闭源2026-04-08
76Claude Opus 4.7 (Non-reasoning, high)历史Anthropic30.93*1,071.03
裁判协议 v4.3.1
75.7%44.7%54.1%33.3%5.1%闭源2026-04-16
77DeepSeek V4 Pro (max)历史DeepSeek30.871,223.37
裁判协议 v4.3.1
924.56
裁判协议 v4.3.1
56.3%14.6%50.8%74.7%43%94.1%13.4%37.5%12.9%开放2026-04-24
78K2 Horizon 375B A23BInstitute of Foundation Models30.791,400
裁判协议 v4.3.1
1,298.34
裁判协议 v4.3.1
37.2%1.5%42.9%80%18.2%25.9%7.4%32%4.6%开放2026-09-03
79GPT-5.5 (low)历史OpenAI30.71*1,112.71
裁判协议 v4.3.1
81%54.9%88%32.7%8%闭源2026-04-23
80Kimi K3 (low)Kimi30.481,185.27
裁判协议 v4.3.1
984.78
裁判协议 v4.3.1
49.3%12.6%52.7%79.3%45.8%77.1%18.6%25%3.1%开放2026-07-16
81Claude Sonnet 4.6 (max)历史Anthropic30.451,295.41
裁判协议 v4.3.1
1,065.11
裁判协议 v4.3.1
20.1%3%50.1%80%40.9%48.4%15.8%33.6%3.1%闭源2026-02-17
82GPT-5.2 (xhigh)历史OpenAI30.45*82.7%44.3%81.2%37.7%11.6%闭源2025-12-11
83GPT-5.6 Terra (medium)OpenAI30.421,318.34
裁判协议 v4.3.1
1,040.72
裁判协议 v4.3.1
33.5%1%50.5%74%44.6%89.8%17%33.3%17.4%闭源2026-07-09
84Apodex 1.1Apodex30.37*1,273.14
裁判协议 v4.3.1
45.5%79.3%31.7%78.4%34.1%4.6%闭源2026-08-30
85Gemini 3.1 Pro PreviewGoogle30.36903.94
裁判协议 v4.3.1
458.83
裁判协议 v4.3.1
35.4%4%58.7%82%54.9%50.9%17.8%47%17.7%闭源2026-02-19
86DeepSeek V4 Pro (high)历史DeepSeek30.11*1,218.72
裁判协议 v4.3.1
70.3%41.4%88.7%35.2%10%开放2026-04-24
87Qwen3.7 Max历史Alibaba29.871,189.86
裁判协议 v4.3.1
917.94
裁判协议 v4.3.1
23%1.5%49.5%79%31.1%25.6%8.6%40.5%13.4%闭源2026-05-19
88MiniMax-M3MiniMax29.611,303.96
裁判协议 v4.3.1
1,096.03
裁判协议 v4.3.1
21.3%2%47.1%83%16.7%18.4%9.8%39%3.7%开放2026-06-01
89Claude Opus 4.5历史Anthropic29.1*77.3%46.6%61%30.1%4.6%闭源2025-11-24
90Claude Sonnet 5 (Non-reasoning)Anthropic28.94*1,283.47
裁判协议 v4.3.1
70%33.8%52%19%1.1%闭源2026-06-30
91MiMo-V2-Pro历史Xiaomi28.64*68.3%26.6%30%30.4%0.3%闭源2026-03-18
92GPT-5.2 Codex (xhigh)历史OpenAI28.5*82.3%41.1%73.4%35.7%8.7%闭源2025-12-11
93Claude Sonnet 5 (medium)Anthropic28.441,222.52
裁判协议 v4.3.1
1,053.88
裁判协议 v4.3.1
27.9%2%51.6%73.7%37.1%69.9%11.6%30%8.6%闭源2026-06-30
94Qwen3.6 Max Preview历史Alibaba28.37*80.7%37.9%46.2%30.8%3.7%闭源2026-04-20
95GPT-5.6 Sol (Non-reasoning)OpenAI28.33*1,273.32
裁判协议 v4.3.1
1,015.25
裁判协议 v4.3.1
22.7%47.7%62.3%48.7%92.8%15.2%16.7%5.1%闭源2026-07-09
96Nex-N2-ProNex AGI28.2*1,175.2
裁判协议 v4.3.1
17.6%43.3%81.7%34.8%94.9%12.6%33.7%8.6%开放2026-06-02
97Solar Pro 4Upstage28.15*1,172.77
裁判协议 v4.3.1
9.1%0.5%44.6%74%18.9%24.4%5.6%29.2%5.4%闭源2026-08-06
98Gemini 3 Pro Preview (high)历史Google27.96*76%55.8%91.5%39.7%9.1%闭源2025-11-18
99GPT-5.6 Terra (low)OpenAI27.921,177.99
裁判协议 v4.3.1
1,014.49
裁判协议 v4.3.1
29.1%1.5%49.9%71.3%43.7%89.9%17.6%29.2%9.4%闭源2026-07-09
100GLM-5历史Z AI27.91*75.7%26.3%35.3%29.3%2%开放2026-02-11
101Qwen3.8 27B (medium)Alibaba27.811,430.39
裁判协议 v4.3.1
1,375.7
裁判协议 v4.3.1
40.3%5.1%39%79.7%18.4%66.7%12.8%14.1%0%开放2026-08-14
102Agnes 2.5 Pro AlphaSapiens AI27.81,087.81
裁判协议 v4.3.1
1,195.25
裁判协议 v4.3.1
38.4%3%42.9%75.7%33.5%88.1%9%33.6%10.9%开放2026-07-24
103GPT-5.4 (low)历史OpenAI27.58*76.7%47.9%82.6%30.8%7.4%闭源2026-03-05
104Kimi K2.6历史Kimi27.461,096.33
裁判协议 v4.3.1
823.87
裁判协议 v4.3.1
13%0.5%51.5%81%32.6%40.5%13%37.5%8%开放2026-04-20
105JT-4.1 Flash 236B A21BChina Mobile27.29*1,193.3
裁判协议 v4.3.1
71.3%24.3%43.7%17.8%0%闭源2026-07-09
106Grok Build 0.1 0616历史SpaceXAI27.17*1,138.29
裁判协议 v4.3.1
74.7%51.5%93.1%38.3%9.1%闭源2026-06-16
107Quasar 438B (max)Multiverse Computing27.131,239.08
裁判协议 v4.3.1
1,123.48
裁判协议 v4.3.1
23.9%1%48.1%76.3%15.5%21.4%5.4%18.7%9.4%闭源2026-08-10
108Qwen3.6 Plus历史Alibaba27.01*1,065.83
裁判协议 v4.3.1
78.3%26.4%34.6%27.8%2.9%闭源2026-04-02
109GPT-5.5 Instant (June 2026)OpenAI26.8661.83
裁判协议 v4.3.1
1,081.76
裁判协议 v4.3.1
36.4%12.6%52.5%70%44.4%73.3%20.2%19.9%0%闭源2026-06-25
110GLM-5-Turbo历史Z AI26.6*71.7%28.4%62.6%27.8%0.3%闭源2026-03-15
111GPT-5.2 (medium)历史OpenAI26.5*70.3%38.3%61.6%26.7%7.9%闭源2025-12-11
112Qwen3.8 27B (low)Alibaba26.491,402.54
裁判协议 v4.3.1
1,277.44
裁判协议 v4.3.1
30.1%2.5%40%77.3%17.1%52.9%10.4%14%0%开放2026-08-14
113GLM-5.1历史Z AI26.451,180.01
裁判协议 v4.3.1
964.01
裁判协议 v4.3.1
20.3%2%44.8%73.7%23.7%29.9%8.4%30.1%4.6%开放2026-04-07
114MiMo-V2.5-ProXiaomi26.411,186.32
裁判协议 v4.3.1
883.43
裁判协议 v4.3.1
13.7%0%50.6%79.7%22.4%24.7%4%35.7%4%开放2026-04-22
115Claude Opus 4.6 (Non-reasoning, high)历史Anthropic26.35*67%45.8%80.1%19.1%2.8%闭源2026-02-05
116Gemini 3 Flash历史Google26.33*78%53.4%93%36.6%8.6%闭源2025-12-17
117Kimi K2.7 CodeKimi26.271,114.49
裁判协议 v4.3.1
855.7
裁判协议 v4.3.1
24.5%1%47.8%79.3%39.6%82.4%11.2%35%10%开放2026-06-12
118Inkling SmallThinking Machines26.091,191.44
裁判协议 v4.3.1
912.02
裁判协议 v4.3.1
4.8%1%49.7%75.7%33.2%63%11.2%33.3%8.3%开放2026-07-30
119Qwen3.7 PlusAlibaba25.82885.83
裁判协议 v4.3.1
912.73
裁判协议 v4.3.1
17.4%1%46.1%73%22.5%27.7%12.2%35.6%9.1%闭源2026-06-01
120Hy3Tencent25.771,135.96
裁判协议 v4.3.1
949.65
裁判协议 v4.3.1
18.4%0.5%48.6%79%32%74.1%11.2%33.5%4.9%开放2026-07-06
121K2 Horizon MoVA 36B A4BInstitute of Foundation Models25.671,243.11
裁判协议 v4.3.1
1,170.41
裁判协议 v4.3.1
26.1%0%40%71.7%15%30.5%4.2%23.4%2.3%开放2026-09-03
122Grok 4.20 0309 v2历史SpaceXAI25.66*69%27.5%17.4%34.5%6.6%闭源2026-04-07
123InklingThinking Machines25.541,165.19
裁判协议 v4.3.1
835.63
裁判协议 v4.3.1
5%1%47%77.3%41.5%67.7%12.8%31.9%5.4%开放2026-07-15
124GPT-5.6 Luna (medium)OpenAI25.481,164.13
裁判协议 v4.3.1
934.87
裁判协议 v4.3.1
23.1%0.5%46.8%75%40.7%90.9%15.2%25.8%4.9%闭源2026-07-09
125Grok 4.3 (high)历史SpaceXAI25.391,018.49
裁判协议 v4.3.1
761.59
裁判协议 v4.3.1
1.4%0%48.3%73%34.8%25.8%5.8%37.2%8%闭源2026-04-30
126Grok 4.20 0309历史SpaceXAI25.22*67.7%28.9%22.4%32.4%6%闭源2026-03-10
127MiMo-V2-Omni-0327历史Xiaomi25.12*73.3%17.8%40.1%22.6%0.9%闭源2026-03-27
128Ling-3.0-flash-VLInclusionAI24.971,224.84
裁判协议 v4.3.1
985.98
裁判协议 v4.3.1
15.7%0%44.2%78.3%14.3%22%8.4%22%2%开放2026-09-10
129GPT-5 Codex (high)历史OpenAI24.88*71.7%38.7%75.4%27.8%5.1%闭源2025-09-23
130DeepSeek V4 Flash (high)历史DeepSeek24.841,082.61
裁判协议 v4.3.1
1,052.09
裁判协议 v4.3.1
28.5%3%40.2%72%34.9%89%10.4%30.3%3.4%开放2026-04-24
131Grok 4.3 (medium)历史SpaceXAI24.78*75%28.8%16.9%30%4.9%闭源2026-04-30
132Claude Sonnet 5 (low)Anthropic24.741,145.04
裁判协议 v4.3.1
927.41
裁判协议 v4.3.1
19.9%2.5%50.1%67.3%37.4%72.8%9.4%21.9%4.6%闭源2026-06-30
133Solar Open2 250BUpstage24.74*1,052.82
裁判协议 v4.3.1
48%71.7%18.8%25.4%28.5%5.7%开放2026-08-12
134GPT-5.1 (high)历史OpenAI24.74*929.53
裁判协议 v4.3.1
80%37.7%51.9%28.5%4.9%闭源2025-11-13
135Claude Sonnet 4.6 (Non-reasoning, high)历史Anthropic24.69*68.3%38.6%68.5%13.3%0.9%闭源2026-02-17
136DeepSeek V4 Flash (max)历史DeepSeek24.631,115.64
裁判协议 v4.3.1
833.07
裁判协议 v4.3.1
28%2.5%45.3%74.3%36.8%96.1%10.8%34.8%7.1%开放2026-04-24
137GPT-5.4 mini (xhigh)历史OpenAI24.581,094.54
裁判协议 v4.3.1
719.24
裁判协议 v4.3.1
26.2%2%52.1%77%37.5%90.2%13.4%28.1%10%闭源2026-03-17
138Grok 4.3 (low)历史SpaceXAI24.3*74%26.7%17.4%18.4%0.6%闭源2026-04-30
139GLM-5.1 (Non-reasoning)历史Z AI24.24*53.3%25.2%63.6%27.9%0%开放2026-04-07
140MiMo-V2-Omni历史Xiaomi23.92*75%19.3%48.9%22.1%1.1%闭源2026-03-19
141Gemini 3.5 Flash (minimal)历史Google23.85*61.3%43%74.3%24.1%1.4%闭源2026-05-19
142GPT-5.1 Codex (high)历史OpenAI23.7*69.3%39.9%77.2%25.7%5.7%闭源2025-11-13
143Claude Opus 4.5 (Non-reasoning)历史Anthropic23.68*70.7%40.9%76.2%13.2%0.3%闭源2025-11-24
144Kimi K2.6 (Non-reasoning)历史Kimi23.57*69.7%24.3%44.2%19.6%1.4%开放2026-04-20
145GLM 5V Turbo历史Z AI23.5*70.3%29.3%68.8%17.1%0.6%闭源2026-04-01
146Kimi K2.5历史Kimi23.46*936.32
裁判协议 v4.3.1
78%35.2%65.7%30.7%3.1%开放2026-01-27
147Nemotron 3 UltraNVIDIA23.411,090.88
裁判协议 v4.3.1
878.22
裁判协议 v4.3.1
3%0.5%40.3%79.3%22.6%29.7%5%28.4%3.1%开放2026-06-04
148Claude Sonnet 4.6 (Non-reasoning, low)历史Anthropic23.3*69.3%36.5%60.8%11.2%0.9%闭源2026-02-17
149MiniMax-M2.7历史MiniMax23.221,087.02
裁判协议 v4.3.1
715.25
裁判协议 v4.3.1
4.4%0%50.1%78.3%26.8%35.6%6%29.6%0.6%开放2026-03-18
150GPT-5.5 (Non-reasoning)历史OpenAI23.17*1,054.41
裁判协议 v4.3.1
64%45.6%92.7%13.7%1.4%闭源2026-04-23
151Ling-3.0-flash-FinInclusionAI23.021,170.55
裁判协议 v4.3.1
967.05
裁判协议 v4.3.1
7.4%0%42.4%73.7%17.9%39.5%5.4%22.6%2.6%开放2026-09-11
152GPT-5 (high)历史OpenAI22.98*1,014.92
裁判协议 v4.3.1
78.2%40.3%82.2%28.5%5.7%闭源2025-08-07
153Qwen3.5 27B历史Alibaba22.9*77.7%20.7%81.5%23.9%0.9%开放2026-02-24
154GPT-5 (medium)历史OpenAI22.87*76%39.5%83.2%25.4%0%闭源2025-08-07
155MiniMax-M2.5历史MiniMax22.8*73.3%26.2%88.1%20.5%1.1%开放2026-02-12
156Hy3-preview历史Tencent22.74*64.7%27.9%87.4%27.8%4.6%开放2026-04-23
157A.X-K2SK Telecom22.7*1,044.19
裁判协议 v4.3.1
41%70%18.6%33%29.6%8.9%开放2026-08-12
158GPT-5.5 Instant (May 2026)历史OpenAI22.69*66.3%46.3%65.3%21.6%2.6%闭源2026-05-05
159Gemini 3.5 Flash-LiteGoogle22.661,063.25
裁判协议 v4.3.1
645.32
裁判协议 v4.3.1
25%1%41.3%76%29.5%34.4%13.6%18.8%0%闭源2026-07-21
160Grok 4历史SpaceXAI22.49*68%40.5%64.5%26.7%2%闭源2025-07-10
161MiMo-V2-Flash (Feb 2026)Xiaomi22.44*71.3%19.8%48.4%22.1%2.9%开放2025-12-16
162GLM-5.2 (Non-reasoning)历史Z AI22.43*1,306.43
裁判协议 v4.3.1
42.3%20.3%33.7%9.8%3.1%开放2026-06-16
163Qwen3.8 27BAlibaba22.42*1,113.3
裁判协议 v4.3.1
12.4%36.2%69.3%8.6%18.1%12.1%0.3%开放2026-08-14
164Gemini 3 Pro Preview (low)历史Google22.32*74%48.3%90%29.5%0%闭源2025-11-18
165MiMo-V2.5Xiaomi22.31,079.43
裁判协议 v4.3.1
750.72
裁判协议 v4.3.1
18.4%0%43.9%73%16.8%31.9%4%27.2%3.7%开放2026-04-22
166GLM-4.7历史Z AI22.24*1,095.87
裁判协议 v4.3.1
71%29.3%93%27.4%1.7%开放2025-12-22
167Kimi K2 Thinking历史Kimi22.02*72%30.9%75.6%23.8%2.6%开放2025-11-06
168Qwen3.6 27B历史Alibaba21.911,068.87
裁判协议 v4.3.1
812.14
裁判协议 v4.3.1
11%0%42.8%77.3%19.6%49.3%11%23.1%1.1%开放2026-04-22
169G9v3-39A5BAI9Stars21.83*1,188.86
裁判协议 v4.3.1
36.8%65.3%14.9%13%17.5%0.3%开放2026-08-20
170GLM-5 (Non-reasoning)历史Z AI21.78*43.7%23.1%45.3%7.6%0%开放2026-02-11
171KAT-Coder-Pro V2KwaiKAT21.7*843.25
裁判协议 v4.3.1
73%21.5%56%16.1%0%闭源2026-03-27
172GPT-5.6 Luna (low)OpenAI21.551,075.16
裁判协议 v4.3.1
743.81
裁判协议 v4.3.1
11.7%0%46.1%70%39.6%89.9%14.2%19.8%2.6%闭源2026-07-09
173DeepSeek V3.2历史DeepSeek21.49*810.24
裁判协议 v4.3.1
73.3%33%82.7%24.6%2.9%开放2025-12-01
174Qwen3.5 397B A17B (Non-reasoning)Alibaba21.44*64.3%24.5%82.7%19.8%0.9%开放2026-02-16
175Qwen3 Max Thinking历史Alibaba21.26*74.3%29.8%94.7%28%1.7%闭源2026-01-26
176GPT-5.4 nano (xhigh)历史OpenAI21.241,035.42
裁判协议 v4.3.1
669.58
裁判协议 v4.3.1
6%0.5%47.2%76.7%25.7%74.2%7.8%28.3%9.3%闭源2026-03-17
177GPT-5.6 Terra (Non-reasoning)OpenAI21.211,145.71
裁判协议 v4.3.1
947.67
裁判协议 v4.3.1
12.5%0.5%45.1%58.7%36.8%95%11.2%11.4%2%闭源2026-07-09
178Claude 4.5 Sonnet历史Anthropic21.18989.26
裁判协议 v4.3.1
711.96
裁判协议 v4.3.1
14.1%0%45.7%72.3%32.9%49.1%5.2%17.8%1.1%闭源2025-09-29
179K2 Horizon 7BInstitute of Foundation Models21.051,091.08
裁判协议 v4.3.1
1,020.86
裁判协议 v4.3.1
20.9%1%27.5%69.7%12.8%33.6%3.2%18.2%0.3%开放2026-09-03
180MiniMax-M2.1历史MiniMax20.95*67.7%21.2%68.5%23.2%0.3%开放2025-12-23
181DeepSeek V4 Pro (Non-reasoning)历史DeepSeek20.83*53%30.9%87.8%8.2%0.9%开放2026-04-24
182MiMo-V2-Flash历史Xiaomi20.82*70.7%24.8%93.4%22.8%4.3%开放2025-12-16
183GPT-5 (low)历史OpenAI20.79*38.1%79%19.6%1.1%闭源2025-08-07
184Ling 3.0 FlashInclusionAI20.631,017.28
裁判协议 v4.3.1
802.47
裁判协议 v4.3.1
3.2%0%42%73%18.2%44.1%5.4%23.7%1.7%开放2026-08-04
185GPT-5 mini (medium)历史OpenAI20.6*72.3%23.1%44.3%15.9%1.4%闭源2025-08-07
186Qwen3.5 Omni PlusAlibaba20.38*63.7%17.4%37%14.9%0.6%闭源2026-03-30
187GPT-5.1 Codex mini (high)历史OpenAI20.38*66.7%23.2%51.6%18.5%0%闭源2025-11-13
188Grok 4.1 Fast历史SpaceXAI20.37*74%25.1%73.4%19.3%2.9%闭源2025-11-19
189o3OpenAI20.2*74.7%38.6%88.1%20.1%1.1%闭源2025-04-16
190GPT-5.4 nano历史OpenAI20.01*67.3%21.9%51.1%15.9%5.1%闭源2026-03-17
191Qwen3.6 27B (Non-reasoning)历史Alibaba19.83*1,042.8
裁判协议 v4.3.1
66.7%16.7%84%15.1%0.9%开放2026-04-22
192GPT-5.4 mini (medium)历史OpenAI19.75*67%36.5%89.8%18.6%2.9%闭源2026-03-17
193K-EXAONE 2.0LG AI Research19.69*918
裁判协议 v4.3.1
42%60%13.1%22.6%18.6%0.9%开放2026-08-12
194LongCat 2.0LongCat19.69964.51
裁判协议 v4.3.1
738.88
裁判协议 v4.3.1
4.5%0%36.3%65%29.6%75.4%6.4%33.7%2.6%开放2026-06-29
195Step 3.7 FlashStepFun19.48*954.2
裁判协议 v4.3.1
43.9%73.7%25.8%85%21.4%2.3%开放2026-05-29
196Kimi K2.5 (Non-reasoning)历史Kimi19.43*67.3%24.1%50%13.2%0.6%开放2026-01-27
197Qwen3.5 27B (Non-reasoning)历史Alibaba19.37*64%15.7%75.1%13.9%0.3%开放2026-02-24
198Claude 4.5 Sonnet (Non-reasoning)历史Anthropic19.34*54%28.4%52.7%7.2%0%闭源2025-09-29
199Qwen3.5 35B A3B历史Alibaba19.33*72%20.1%85.4%21%0.9%开放2026-02-24
200Qwen3.5 397B A17BAlibaba19.11905.4
裁判协议 v4.3.1
562.65
裁判协议 v4.3.1
6.9%0%44.8%77.3%30.8%88.9%12.2%29%1.7%开放2026-02-16
201Claude 4 Sonnet历史Anthropic18.92*818.93
裁判协议 v4.3.1
70.3%22.7%29.1%10.7%0.3%闭源2025-05-22
202DeepSeek V4 Flash (Non-reasoning)历史DeepSeek18.88*41.7%26%94.5%7.8%0.3%开放2026-04-24
203Qwen3.6 35B A3BAlibaba18.81990.51
裁判协议 v4.3.1
711.46
裁判协议 v4.3.1
5.2%0%36.6%71.7%18.8%50.5%6.6%22.2%0.3%开放2026-04-16
204JT-35B-FlashChina Mobile18.66*65.7%24.2%63.8%6.4%0%闭源2026-05-14
205MiniMax-M2历史MiniMax18.63*64.3%23.2%90.9%13.7%0.9%开放2025-10-26
206KAT-Coder-Pro V1历史KwaiKAT18.59*845.3
裁判协议 v4.3.1
602.09
裁判协议 v4.3.1
38.7%17.7%67.4%33.6%0%闭源2025-11-11
207GLM-4.6历史Z AI18.53*873.8
裁判协议 v4.3.1
54%26.9%94.1%14.5%1.1%开放2025-09-30
208MiMo-V2.5-Pro (Non-reasoning)Xiaomi18.29*41.7%27.2%89.2%14.8%1.1%开放2026-04-22
209GPT-5.4 (Non-reasoning)历史OpenAI18.16*58.3%37.4%84.7%11.3%0.6%闭源2026-03-05
210Muse Glimmer (high)Meta18.07892.72
裁判协议 v4.3.1
481.02
裁判协议 v4.3.1
6.8%0.5%44.9%83.3%27%81.9%10%22%2.6%开放2026-08-10
211Grok 4 Fast历史SpaceXAI17.94*73.7%22.8%68.3%19.1%2.9%闭源2025-09-19
212Gemini 3 Flash (Non-reasoning)历史Google17.93*55.3%45.8%92.4%15%1.4%闭源2025-12-17
213Qwen3.5 122B A10B (Non-reasoning)Alibaba17.75*833.05
裁判协议 v4.3.1
61.3%19.1%91.6%15.9%0.9%开放2026-02-24
214Claude 3.7 Sonnet历史Anthropic17.71*28%40%9.7%0.9%闭源2025-02-24
215Claude 4.5 HaikuAnthropic17.59854.45
裁判协议 v4.3.1
614.13
裁判协议 v4.3.1
3.2%0%42.2%74.3%18%27.3%3.8%10.4%0%闭源2025-10-15
216GPT-5 mini (high)历史OpenAI17.41880.27
裁判协议 v4.3.1
437.58
裁判协议 v4.3.1
6.5%0%39%72.3%25%56.4%8.4%21.5%0%闭源2025-08-07
217GLM-4.7 (Non-reasoning)历史Z AI17.36*40.7%23.6%92.9%6.4%0%开放2025-12-22
218Ring-2.6-1TInclusionAI17.28863.62
裁判协议 v4.3.1
654.34
裁判协议 v4.3.1
0.7%0.5%45%70%25.7%85.3%3.8%21.6%3.7%开放2026-05-08
219Hy3-preview (Non-reasoning)历史Tencent17.04*42%23.2%75.7%7%0.3%开放2026-04-23
220Ling-2.6-1T历史InclusionAI17*41.7%21.9%92.8%8.7%0.3%开放2026-04-23
221GPT-5.2 (Non-reasoning)历史OpenAI16.98*45.7%30.9%62.3%8%0.6%闭源2025-12-11
222Step 3.5 Flash 2603历史StepFun16.96*63%24.6%91.3%24.5%2.3%闭源2026-04-02
223Doubao Seed CodeByteDance Seed16.95*71.7%24.9%80%14.1%0.3%闭源2025-11-11
224Gemma 4 26B A4BGoogle16.67*693
裁判协议 v4.3.1
40%65.7%19.1%86.4%19.3%0%开放2026-04-02
225Gemini 2.5 Pro历史Google16.66615.9
裁判协议 v4.3.1
315.4
裁判协议 v4.3.1
2.2%0%46.3%69%39.1%90.9%10.2%22.5%2.6%闭源2025-06-05
226o4-mini (high)历史OpenAI16.65*61%24.8%80.5%16.5%0.6%闭源2025-04-16
227Step 3.5 Flash历史StepFun16.62*50%23.6%85.7%21.1%2.5%开放2026-02-02
228Claude 4 Sonnet (Non-reasoning)历史Anthropic16.61*44%22.7%41%4.3%1.1%闭源2025-05-22
229DeepSeek V3.2 Exp历史DeepSeek16.58*72.3%27.6%80.8%14.9%1.4%开放2025-09-29
230Qwen3 Max Thinking (Preview)历史Alibaba16.29*62%27.4%91.5%12.7%0%闭源2025-11-03
231Qwen3.5 122B A10BAlibaba16.19925.2
裁判协议 v4.3.1
498.94
裁判协议 v4.3.1
5.5%0%39.7%76.3%24.4%87.1%3.4%25.2%0.6%开放2026-02-24
232K2 Horizon 3.7BInstitute of Foundation Models16.16962.39
裁判协议 v4.3.1
792.8
裁判协议 v4.3.1
14.7%0%22%62.3%10.8%44.1%3.2%13.9%0%开放2026-09-03
233GPT-5.6 Luna (Non-reasoning)OpenAI16.05987.47
裁判协议 v4.3.1
710.24
裁判协议 v4.3.1
5%1%40.4%42.7%28.6%75%6%7.2%0.3%闭源2026-07-09
234MiMo-V2-Flash (Non-reasoning)历史Xiaomi16.05*788.18
裁判协议 v4.3.1
36.3%15.6%76%8.6%0%开放2025-12-16
235DeepSeek V3.2 (Non-reasoning)历史DeepSeek16.04*45.7%24%93.3%11.2%0.9%开放2025-12-01
236Gemini 3.1 Flash-Lite历史Google16.03595.99
裁判协议 v4.3.1
226.15
裁判协议 v4.3.1
6.8%0.5%43.4%74.3%36.3%82.7%7.8%17.2%1.1%闭源2026-03-03
237Qwen3 Max历史Alibaba15.61*50%24.4%89.9%11.9%0%闭源2025-09-23
238Gemini 2.5 Flash (Sep)历史Google15.47*71%28.3%89.9%13.8%0.3%闭源2025-09-25
239Gemma 4 31BGoogle15.42755.45
裁判协议 v4.3.1
376.57
裁判协议 v4.3.1
4.9%0%45.5%69.7%20%85%6%23.6%1.4%开放2026-04-02
240Claude 4.5 Haiku (Non-reasoning)Anthropic15.41*49.7%14.4%25.7%4.2%0%闭源2025-10-15
241DeepSeek V3.1 Terminus历史DeepSeek15.4826.12
裁判协议 v4.3.1
464.01
裁判协议 v4.3.1
2.7%0%38%69.3%27.7%74.8%5.4%16.4%1.7%开放2025-09-22
242Kimi K2 0905历史Kimi15.29*53%25.4%69.7%6.4%0%开放2025-09-05
243Claude 3.7 Sonnet (Non-reasoning)历史Anthropic15.27*51.7%28.2%52.8%4.2%0%闭源2025-02-24
244Qwen3.6 35B A3B (Non-reasoning)Alibaba15.23*953.19
裁判协议 v4.3.1
64.3%16.7%92.1%13.9%0%开放2026-04-16
245o1历史OpenAI15.23*65%34.5%69.6%7%0.3%闭源2024-12-05
246Qwen3.5 35B A3B (Non-reasoning)历史Alibaba15.12*740.47
裁判协议 v4.3.1
63%15.7%93.4%13.4%0.6%开放2026-02-24
247GLM-4.6 (Non-reasoning)历史Z AI14.93*26.3%21.4%67.6%5.5%0%开放2025-09-30
248Mistral Medium 3.5Mistral14.89875.14
裁判协议 v4.3.1
523.63
裁判协议 v4.3.1
6.3%0%40.2%69.3%24.7%81.6%2.8%13.8%0%开放2026-04-29
249GLM-4.7-Flash历史Z AI14.87*41.7%16.2%93.9%7.6%0.3%开放2026-01-19
250Granite 4.2 30BIBM14.83*721.01
裁判协议 v4.3.1
37.8%49%10.1%25.6%1.8%11.2%0.3%开放2026-08-25
251Grok 3 mini Reasoning (high)历史SpaceXAI14.62*55.3%15.1%25.8%11%0.6%闭源2025-02-19
252Grok 4.20 0309 (Non-reasoning)历史SpaceXAI14.61*25.3%26.1%92.8%24.5%0%闭源2026-03-10
253Grok 4.3 (Non-reasoning)历史SpaceXAI14.541,030.8
裁判协议 v4.3.1
724.4
裁判协议 v4.3.1
0.9%0%39.4%32.3%23.7%74.5%2.8%6.8%0%闭源2026-04-30
254DeepSeek V3.2 Speciale历史DeepSeek14.46*70%37.8%89.2%28.7%7.4%开放2025-12-01
255K-EXAONE历史LG AI Research14.37*539.71
裁判协议 v4.3.1
61.3%16.4%88.8%13.9%1.1%开放2025-12-31
256ERNIE 5.0 Thinking PreviewBaidu14.26*22.1%87%13.3%1.4%闭源2025-11-13
257Grok 4.20 0309 v2 (Non-reasoning)历史SpaceXAI14.2*23.3%26.7%97.3%27.9%0.3%闭源2026-04-07
258Gemma 4 12BGoogle14.18*590.73
裁判协议 v4.3.1
63.7%15.6%81%15.7%0%开放2026-06-03
259Nova 2.0 Pro Preview (medium)Amazon14.16*629.57
裁判协议 v4.3.1
64%22%91%9.4%0%闭源2025-11-27
260Grok Code Fast 1历史SpaceXAI14.06*53%23.5%79.3%8%0%闭源2025-08-28
261DeepSeek V3.1 Terminus (Non-reasoning)历史DeepSeek13.93*45.3%23.5%87.6%8.7%0%开放2025-09-22
262Command A+Cohere13.92658.08
裁判协议 v4.3.1
366.16
裁判协议 v4.3.1
1.8%0.5%38.5%52.7%8.9%14.2%1%12%0.3%开放2026-05-20
263Gemma 4 31B (Non-reasoning)Google13.92*691.46
裁判协议 v4.3.1
46.7%16.6%81.9%11.8%0%开放2026-04-02
264DeepSeek V3.2 Exp (Non-reasoning)历史DeepSeek13.88*44.7%22.8%92.1%9%1.4%开放2025-09-29
265Apriel-v1.5-15B-Thinker历史ServiceNow13.82*21.3%16.4%85.3%12.1%1.1%开放2025-09-30
266DeepSeek V3.1 (Non-reasoning)历史DeepSeek13.71*47%23.1%85.7%6.7%0%开放2025-08-21
267Qwen3.5 9BAlibaba13.65*588.42
裁判协议 v4.3.1
70%16.4%83.6%14.9%0.3%开放2026-03-02
268Nova 2.0 Omni (medium)Amazon13.65*59.7%17.7%93.3%7%0%闭源2025-11-26
269Nemotron 3.5 LightningNVIDIA13.64766.89
裁判协议 v4.3.1
508.5
裁判协议 v4.3.1
0.8%0.5%32.1%60.3%14.4%37.6%3%10.6%0%开放2026-08-11
270Nemotron 3 SuperNVIDIA13.55644.41
裁判协议 v4.3.1
0
裁判协议 v4.3.1
3.8%0%36.2%65.7%24.3%87%2.6%20.8%3.1%开放2026-03-11
271DeepSeek V3.1历史DeepSeek13.48*56.7%29%82.5%14.3%2%开放2025-08-21
272Qwen3 VL 235B A22B (Reasoning)历史Alibaba13.44*63.7%20.8%85.2%11.9%0%开放2025-09-23
273Apriel-v1.6-15B-ThinkerServiceNow13.4*17%91.9%10.8%0.3%开放2025-11-25
274Nova 2.0 Lite (high)Amazon13.37*535.99
裁判协议 v4.3.1
60.3%19%91.4%11.6%0.3%闭源2025-10-29
275GPT-5.1 (Non-reasoning)历史OpenAI13.31*45%29.5%90.6%5.3%0%闭源2025-11-13
276Qwen3.5 9B (Non-reasoning)Alibaba13.27*46%14%98.4%9.4%0.6%开放2026-03-02
277EXAONE 4.5 33BLG AI Research13.21*627.65
裁判协议 v4.3.1
54.7%16.2%79.9%12.9%0.3%开放2026-04-09
278MiniCPM5-2BOpenBMB13.14827.51
裁判协议 v4.3.1
433.52
裁判协议 v4.3.1
1.8%0%26.3%59.3%8.4%21.9%1.4%8.9%0.3%开放2026-09-07
279Gemma 4 26B A4B (Non-reasoning)Google13.13*42.3%15.5%92.1%11.5%0%开放2026-04-02
280Qwen3.5 4BAlibaba13.12*63%15.1%86.6%9.9%0%开放2026-03-02
281DeepSeek R1 0528历史DeepSeek13.12*55.7%30.5%83.4%15.8%1.4%开放2025-05-28
282Gemini 2.5 Flash历史Google13.11*65.3%25.9%75.3%12.1%1.1%闭源2025-05-20
283GPT-5 nano (high)历史OpenAI12.99*45%19.1%59%9.5%0%闭源2025-08-07
284Nova 2.0 Pro Preview (low)Amazon12.78*600.77
裁判协议 v4.3.1
64.7%17.1%69.5%5.2%0%闭源2025-11-27
285GLM-4.5历史Z AI12.77*52.7%25.1%70.1%13%0%开放2025-07-28
286Kimi K2历史Kimi12.72*53%27.4%76.6%7.4%0%开放2025-07-11
287Qwen3 235B A22B 2507历史Alibaba12.71495.33
裁判协议 v4.3.1
189.93
裁判协议 v4.3.1
2.2%0%41.4%72%22.8%89.8%4.8%15.9%0%开放2025-07-25
288GPT-4.1历史OpenAI12.69*68.3%27.8%93.3%4.2%0%闭源2025-04-14
289Qwen3 Max (Preview)历史Alibaba12.59*43%24.2%88.4%10.1%0.9%闭源2025-09-05
290Nova 2.0 Lite (medium)Amazon12.48*60%18.2%91.8%9%0%闭源2025-10-29
291GPT-5 nano (medium)历史OpenAI12.48*43.7%17.6%52.6%8.7%0%闭源2025-08-07
292Qwen3.5 Omni FlashAlibaba12.47*52%14.6%93.4%7.6%0%闭源2026-03-30
293Gemini 2.5 Flash (Sep) (Non-reasoning)历史Google12.38*60%26.8%91.2%8.7%0%闭源2025-09-25
294gpt-oss-120b (high)OpenAI12.35744.74
裁判协议 v4.3.1
0
裁判协议 v4.3.1
0.2%0%34%52%21.8%90.8%4%19.6%1.1%开放2025-08-05
295JT-MINIChina Mobile12.21*13%14.2%91.3%6.4%0%闭源2026-04-15
296Grok 3历史SpaceXAI12.11*58%28.9%87.3%4.1%0%闭源2025-02-19
297Seed-OSS-36B-Instruct历史ByteDance Seed12.1*61.3%18.2%86.4%9.9%0%开放2025-08-20
298Qwen3 235B 2507历史Alibaba12*33.9%18.7%77.2%11.1%0%开放2025-07-21
299Qwen3 Coder 480B历史Alibaba11.9*45.7%15.7%44.9%4.5%0%开放2025-07-22
300Qwen3 VL 32B (Reasoning)历史Alibaba11.87*55.3%17%83.7%10.1%0%开放2025-10-21
301Ling 3.0 TinyInclusionAI11.87715.64
裁判协议 v4.3.1
481.43
裁判协议 v4.3.1
0.3%0%24.2%60.3%8.5%30.5%0.8%9.3%0%开放2026-08-06
302Magistral Medium 1.2Mistral11.83*361.94
裁判协议 v4.3.1
53%20.8%60%10.3%0.3%闭源2025-09-18
303Granite 4.2 8BIBM11.81645.07
裁判协议 v4.3.1
320.12
裁判协议 v4.3.1
2.1%0%31.5%45%11.2%32%0.6%9.7%0.3%开放2026-08-25
304Nova 2.0 Lite (low)Amazon11.8*54.3%17.6%87.1%4%0%闭源2025-10-29
305HyperNova 60B 2605 (high)Multiverse Computing11.74*603.87
裁判协议 v4.3.1
249.39
裁判协议 v4.3.1
36.3%15.2%82.3%16.6%0.9%开放2026-05-26
306MiniMax M1 80k历史MiniMax11.73*57.7%22.5%91.5%8.9%0%开放2025-06-17
307GPT-5.4 nano (Non-reasoning)历史OpenAI11.69*680.74
裁判协议 v4.3.1
29.7%14.9%61.5%4.1%0%闭源2026-03-17
308Nemotron Cascade 2 30B A3BNVIDIA11.67*447.27
裁判协议 v4.3.1
39.3%18.3%85.4%12%0.6%开放2026-03-19
309Mercury 2Inception11.51646.34
裁判协议 v4.3.1
279.6
裁判协议 v4.3.1
1.7%0%37.7%43.7%21.2%91.2%3.8%17.1%0.8%闭源2026-02-20
310K2 Think V2Institute of Foundation Models11.5*328.05
裁判协议 v4.3.1
34
裁判协议 v4.3.1
57%18%71.2%10.1%0%开放2025-12-15
311LongCat Flash LiteLongCat11.47*28.3%13.3%96.2%5.8%0%开放2026-01-28
312Mistral Small 4Mistral11.45536.71
裁判协议 v4.3.1
255.96
裁判协议 v4.3.1
1%0%38.8%49.7%21.7%66.5%0%9.9%0.3%开放2026-03-16
313GPT-5 (minimal)历史OpenAI11.45*29.6%90.1%6%0%闭源2025-08-07
314DeepSeek R1 (Jan)历史DeepSeek11.41478.84
裁判协议 v4.3.1
149.56
裁判协议 v4.3.1
0.2%0%38.3%57.7%30.5%90.3%2.4%8.5%0.6%开放2025-01-20
315HyperCLOVA X SEED Think (32B)Naver11.36*13.7%15.4%81.2%5.5%0%开放2025-12-26
316Grok 4.1 Fast (Non-reasoning)历史SpaceXAI11.28*31.3%17.2%82.3%5.1%0%闭源2025-11-19
317GLM-4.6V历史Z AI11.22*48.7%16.2%51.5%9.6%0%开放2025-12-08
318K-EXAONE (Non-reasoning)历史LG AI Research11.21*53.3%13.7%91.5%5.7%0%开放2025-12-31
319Qwen3 Next 80B A3B (Reasoning)Alibaba11.2*324.88
裁判协议 v4.3.1
63.7%19%87%12.6%0%开放2025-09-11
320GPT-5.4 mini (Non-reasoning)历史OpenAI11.14*734.27
裁判协议 v4.3.1
37%25.7%95.6%5.9%0%闭源2026-03-17
321Nova 2.0 Omni (low)Amazon11.11*57.7%18.6%84.7%0%闭源2025-11-26
322GLM-4.5-Air历史Z AI11.09*46.7%16.3%92.9%7%0%开放2025-07-28
323Grok 4 Fast (Non-reasoning)历史SpaceXAI11.06*24%17.7%87.6%4.5%0%闭源2025-09-19
324Mi:dm K 2.5 ProKorea Telecom11.04*9%17.9%93.9%8.1%0%闭源2025-12-11
325o3-mini (high)历史OpenAI10.96420.79
裁判协议 v4.3.1
161.06
裁判协议 v4.3.1
0.5%0%42.8%43%21.3%81.1%2%12%0.3%闭源2025-01-31
326Ring-1T历史InclusionAI10.9*54.7%20.5%89.8%11.1%0.6%开放2025-10-13
327Trinity Large ThinkingArcee AI10.88511.58
裁判协议 v4.3.1
271.11
裁判协议 v4.3.1
1.3%0.5%40.6%38%22.5%85.9%1.2%15.8%0.9%开放2026-04-01
328G9v3-3BAI9Stars10.84*807.38
裁判协议 v4.3.1
41.3%6.5%11.7%4.5%0%开放2026-07-23
329Qwen3.5 4B (Non-reasoning)Alibaba10.81*34.7%11.2%96.8%8%0%开放2026-03-02
330INTELLECT-3Prime Intellect10.61*39.7%19.6%87.6%13.1%0.3%开放2025-11-27
331GLM-4.7-Flash (Non-reasoning)历史Z AI10.56*20.3%13.1%94.3%5%0%开放2026-01-19
332Solar Open 100BUpstage10.36*41%18.5%89.6%10.4%0%开放2025-12-17
333Gemini 2.5 Flash-Lite (Sep)历史Google10.35*64.7%18%87.8%7%0%闭源2025-09-25
334Nemotron 3 Nano Omni 30B A3BNVIDIA10.25*415.52
裁判协议 v4.3.1
39.7%15.2%85.7%4.8%0%开放2026-04-29
335gpt-oss-120b (low)OpenAI10.21*364.46
裁判协议 v4.3.1
46%19.8%91.4%5.9%0%开放2025-08-05
336GPT-4.1 mini历史OpenAI10.16*453.21
裁判协议 v4.3.1
44%20.3%92.7%5%0%闭源2025-04-14
337Qwen3 Coder NextAlibaba10.05664.31
裁判协议 v4.3.1
413.75
裁判协议 v4.3.1
1.1%0%36.2%47%16.2%93.7%2.6%10.1%0%开放2026-02-03
338Nova 2.0 Pro Preview (Non-reasoning)Amazon9.96*508.59
裁判协议 v4.3.1
30%16.9%79.9%3.9%0%闭源2025-11-27
339gpt-oss-20b (low)OpenAI9.95*31%15.1%86.7%5.3%0%开放2025-08-05
340Qwen3 VL 235B A22B历史Alibaba9.94*32.7%20.3%91.6%6.6%0%开放2025-09-23
341North Mini CodeCohere9.91471.64
裁判协议 v4.3.1
234.19
裁判协议 v4.3.1
1.5%0.5%38.8%37.3%18.9%83.2%1.8%11.1%0.3%开放2026-06-09
342GPT-5 mini (minimal)历史OpenAI9.91*39.3%19.2%89.3%5.1%0%闭源2025-08-07
343K2-V2 (high)Institute of Foundation Models9.87*35%18.5%91%10.5%0%开放2025-12-05
344Gemini 2.5 Flash (Non-reasoning)历史Google9.85*49.9%26.1%93%4.7%1.4%闭源2025-05-20
345Qwen3 30B A3B 2507历史Alibaba9.81273.6
裁判协议 v4.3.1
31.59
裁判协议 v4.3.1
0.8%0%33%61.3%16.3%86.5%0.4%10.3%0.3%开放2025-07-30
346DeepSeek V3 0324历史DeepSeek9.72274.71
裁判协议 v4.3.1
91.96
裁判协议 v4.3.1
0.3%0%39%40.7%24.3%85.9%1.6%4.7%0%开放2025-03-25
347Mistral Large 3Mistral9.71588.39
裁判协议 v4.3.1
245.84
裁判协议 v4.3.1
1.5%0%36.6%36%24.9%86%1.2%4.2%0%开放2025-12-02
348Ling 2.6 Flash历史InclusionAI9.68*492.76
裁判协议 v4.3.1
31.3%15.6%96.7%6.3%0%开放2026-04-21
349Qwen3 Next 80B A3BAlibaba9.64*52.7%17.3%92.7%7.6%0%开放2025-09-11
350Tri-21B-think Preview历史Trillion Labs9.59*21.7%10.1%80.5%5.8%0%开放2026-02-10
351Qwen3 Coder 30B A3B历史Alibaba9.59*32.7%16.4%80.3%3.8%0%开放2025-07-31
352Mistral Medium 3.1历史Mistral9.53536.15
裁判协议 v4.3.1
538.05
裁判协议 v4.3.1
4.1%0%32.4%42.7%20.8%84.6%1.2%4.7%0%闭源2025-08-12
353DiffusionGemma 26B A4BGoogle9.51*501.29
裁判协议 v4.3.1
19.7%16.6%90.9%10.8%0.3%开放2026-06-10
354Qwen3 235B历史Alibaba9.5*0%18.5%77.6%11%0%开放2025-04-28
355Qwen3 VL 30B A3B (Reasoning)历史Alibaba9.45*16.9%89.8%8.9%0%开放2025-10-03
356Devstral 2Mistral9.41688.65
裁判协议 v4.3.1
483.02
裁判协议 v4.3.1
3.1%0%32.8%32.3%20.8%85.3%2.4%3.6%0%开放2025-12-09
357Gemma 4 12B (Non-reasoning)Google9.38*35%11.9%73.4%6.3%0%开放2026-06-03
358Gemini 2.5 Flash-Lite (Sep) (Non-reasoning)历史Google9.34*49.7%14.1%66.1%5.1%0%闭源2025-09-25
359Llama 4 MaverickMeta9.3-44.77
裁判协议 v4.3.1
0
裁判协议 v4.3.1
0.2%0%31.7%50%24.9%88.9%0.8%4.9%0%开放2025-04-05
360Motif-2-12.7BMotif Technologies9.19*14.7%15.2%90.6%8.8%0%闭源2025-12-04
361Ling-1T历史InclusionAI9.17*36.7%20.3%94.7%7.3%0.6%开放2025-10-08
362Nova PremierAmazon9.16*34.3%21.3%71.9%4.2%0%闭源2025-04-30
363Granite 4.2 3BIBM9.06389.62
裁判协议 v4.3.1
120.61
裁判协议 v4.3.1
0.4%0%25.3%24.3%9.2%26.3%0.2%6.6%0%开放2026-08-25
364Magistral Medium 1历史Mistral9.05*0%20.3%59.3%9.8%0.3%闭源2025-06-10
365Mistral Medium 3历史Mistral9.05*31.3%18.3%60.9%4%0%闭源2025-05-07
366gpt-oss-20b (high)OpenAI9.04513.93
裁判协议 v4.3.1
0
裁判协议 v4.3.1
0.2%0%38.9%34.7%16%94.1%2%11%1.4%开放2025-08-05
367K2-V2 (medium)Institute of Foundation Models9.02*28%17.9%81.6%4.4%0%开放2025-12-05
368Llama Nemotron Super 49B v1.5NVIDIA9.01*17.2%78.2%7.3%0%开放2025-07-25
369Devstral Medium历史Mistral9.01*32%18.9%62.2%3.8%0%闭源2025-07-10
370Mistral Small 4 (Non-reasoning)Mistral8.99*28.3%16.6%78%3.8%0.3%开放2026-03-16
371Tri-21B-ThinkTrillion Labs8.99*15.7%12.3%85.4%5.9%0.3%开放2026-02-10
372Gemini 2.0 Flash历史Google8.94*31.3%23.5%86.2%4.3%0%闭源2025-02-05
373Claude 3.5 Haiku历史Anthropic8.94*407.74
裁判协议 v4.3.1
27.3%13.2%41.1%3.6%0%闭源2024-10-22
374Llama 3.3 Nemotron Super 49B历史NVIDIA8.93*18.3%17.1%91%5.6%0%开放2025-03-18
375Gemma 4 E4BGoogle8.91*176.92
裁判协议 v4.3.1
32%8.6%30.9%3.8%0.6%开放2026-04-03
376Nemotron 3 NanoNVIDIA8.9438.32
裁判协议 v4.3.1
100.39
裁判协议 v4.3.1
1%0%30.6%38%17.3%83.3%1%11.4%0.9%开放2025-12-15
377Qwen3 4B 2507历史Alibaba8.8*37.3%12.5%83.2%6.2%0%开放2025-08-06
378MiniCPM5-1BOpenBMB8.8*6%2%17%6.5%0%开放2026-05-25
379Sarvam 105B (high)Sarvam8.79*0%17.6%93.4%11%0%开放2026-03-06
380Nova 2.0 Lite (Non-reasoning)Amazon8.74*18.7%14%85.9%2.9%0%闭源2025-10-29
381Devstral Small (May)历史Mistral8.74*32%15.9%86.4%4%0%开放2025-05-21
382MiniCPM5-1B (Non-reasoning)OpenBMB8.69*5%0.1%0.9%4.5%0%开放2026-05-25
383Magistral Small 1.2Mistral8.59*212.7
裁判协议 v4.3.1
0.2%19.3%13.5%90.8%6.4%0.3%开放2025-09-17
384Gemini 2.5 Flash-Lite历史Google8.55*55.7%17.9%77.4%6.8%0%闭源2025-06-17
385DeepSeek V3 (Dec)历史DeepSeek8.49183.26
裁判协议 v4.3.1
99.91
裁判协议 v4.3.1
0.6%0%35.8%29.3%25.4%90%0.8%2.9%0%开放2024-12-26
386GPT-4o (Nov)历史OpenAI8.44*49.3%19.9%37.9%2.4%0%闭源2024-11-20
387Nanbeige4.1-3BNanbeige8.4*-166.42
裁判协议 v4.3.1
0%9.9%58.1%10.9%0%开放2026-02-11
388LFM2.5-2.6BLiquid AI8.39*202.35
裁判协议 v4.3.1
14.4%5.7%4.3%16%6.2%0%开放2026-08-04
389Qwen3 VL 32B历史Alibaba8.39*14.8%91.8%6.8%0%开放2025-10-21
390Devstral Small 2Mistral8.36679.88
裁判协议 v4.3.1
495.98
裁判协议 v4.3.1
3.1%0%32.4%28%15.8%87.3%1%3.5%0%开放2025-12-09
391GLM-4.6V (Non-reasoning)历史Z AI8.35*17%17.4%66.6%3.7%0%开放2025-12-08
392Qwen3 235B (Non-reasoning)历史Alibaba8.29*0%18.6%87.1%4.2%0%开放2025-04-28
393Magistral Small 1历史Mistral8.22*0%18%76.1%7.6%0%开放2025-06-10
394EXAONE 4.0 32BLG AI Research8.18*13.9%85.8%11.4%0%开放2025-07-15
395Qwen3 VL 8B (Reasoning)历史Alibaba8.17*33.3%20.4%91.4%3.8%0.3%开放2025-10-14
396Nova 2.0 Omni (Non-reasoning)Amazon8.15*24.7%13.2%89.3%3.8%0%闭源2025-11-26
397DeepSeek R1 0528 Qwen3 8B历史DeepSeek8.08*15%12.2%86.4%5.9%0%开放2025-05-29
398Qwen3 VL 30B A3B历史Alibaba7.93*15.6%93.7%6.3%0%开放2025-10-03
399Hermes 4 70BNous Research7.91*9.7%23.5%95.4%8.8%0%开放2025-08-27
400DeepSeek R1 Distill Llama 70B历史DeepSeek7.89*10%19.2%81.1%5.1%0%开放2025-01-20
401Falcon-H1R-7BTII UAE7.83*11%14.5%88.8%11%0.3%开放2026-01-04
402GPT-4.1 nano历史OpenAI7.82*12.78
裁判协议 v4.3.1
20.3%13.7%82.6%3.8%0%闭源2025-04-14
403Solar Pro 3历史Upstage7.82446.94
裁判协议 v4.3.1
121.9
裁判协议 v4.3.1
0.5%0%25.5%32.3%18.5%88.2%1.6%10.3%0%闭源2026-04-06
404Ling-flash-2.0历史InclusionAI7.81*17.7%14.1%89.5%6.2%0%开放2025-09-17
405Gemma 4 E2BGoogle7.77*36.05
裁判协议 v4.3.1
16.3%6.6%32.4%4.8%0%开放2026-04-02
406Qwen3 Omni 30B A3B (Reasoning)Alibaba7.76*0%14.6%89%7.5%0%开放2025-09-22
407GPT-4o (Aug)历史OpenAI7.74*41%23.7%58.4%2.3%0%闭源2024-08-06
408Qwen2.5 72B历史Alibaba7.73*17.5%85.5%3.6%0%开放2024-09-19
409Step3 VL 10BStepFun7.69*0%13.3%83.3%10.8%0%开放2026-01-20
410Llama 3.3 70BMeta7.66*49.51
裁判协议 v4.3.1
15.7%18.9%90.2%3.6%0%开放2024-12-06
411Qwen3 30B历史Alibaba7.63*0%16.2%80.7%6.2%0%开放2025-04-28
412Devstral Small历史Mistral7.6*18.7%14.7%76.9%3.8%0%开放2025-07-10
413GLM-4.5V历史Z AI7.56*0%20.8%84.8%6.3%0%开放2025-08-11
414Mistral Large 2 (Nov)历史Mistral7.56*19.9%67.7%3.3%0%开放2024-11-18
415Llama Nemotron UltraNVIDIA7.53*20.1%81.2%7.4%0%开放2025-04-07
416Qwen3 30B A3B 2507 (Non-reasoning)历史Alibaba7.52*26.3%15%95.4%6.9%0%开放2025-07-29
417ERNIE 4.5 300B A47BBaidu7.5*19.1%66.9%3.3%0%开放2025-06-30
418Hermes 4 405BNous Research7.49*22.3%30.1%94.5%10.9%0.3%开放2025-08-27
419Solar Pro 2历史Upstage7.49*0%20%94.7%7.4%0%闭源2025-07-09
420NVIDIA Nemotron Nano 12B v2 VLNVIDIA7.48*41%14.6%92.3%5.5%0%开放2025-10-28
421Gemma 4 E4B (Non-reasoning)Google7.46*24%8.3%53.8%4.8%0.3%开放2026-04-03
422Granite 4.1 30B历史IBM7.44*378.65
裁判协议 v4.3.1
23.7%14%95%4.1%0%开放2026-04-29
423NVIDIA Nemotron Nano 9B V2NVIDIA7.43*22.7%11.8%60.5%4.9%0%开放2025-08-18
424Hermes 4 405B (Non-reasoning)Nous Research7.41*22%26.1%80.2%4.2%0%开放2025-08-27
425Nemotron 3 Nano 4BNVIDIA7.38*159.11
裁判协议 v4.3.1
21.7%9.6%88.6%4.9%0%开放2026-03-16
426Mistral Small 3.1历史Mistral7.37549.93
裁判协议 v4.3.1
324.58
裁判协议 v4.3.1
1.2%0%27.8%22.3%15.1%77.6%1%4.3%0%开放2025-03-17
427Llama Nemotron Super 49B v1.5 (Non-reasoning)NVIDIA7.36*24.7%12.6%67.3%4.3%0%开放2025-07-25
428K2-V2 (low)Institute of Foundation Models7.31*20.3%16.3%76%3.6%0%开放2025-12-05
429Llama 3.1 405BMeta7.29*25.3%23.2%52.4%4%0%开放2024-07-23
430Llama 3.3 Nemotron Super 49B (Non-reasoning)历史NVIDIA7.28*12%13.5%73.7%3.8%0%开放2025-03-18
431Qwen3 VL 8B历史Alibaba7.26*16.7%20.5%91%2.7%0%开放2025-10-14
432LFM2.5-8B-A1BLiquid AI7.22*0%9.3%46.9%6.9%0%开放2026-05-28
433Qwen3 32B历史Alibaba7.22239.65
裁判协议 v4.3.1
9.49
裁判协议 v4.3.1
1%0%36%0%17.4%82.1%1.6%7.4%0.3%开放2025-04-28
434Ring-flash-2.0InclusionAI7.15*21%16.6%90.1%9.6%0.3%开放2025-09-19
435Olmo 3.1 32B ThinkAllen Institute for AI7.12*0%13.8%66.3%6.3%0%开放2025-12-12
436GPT-5 nano (minimal)历史OpenAI7.09*20%13.1%88.8%4%0%闭源2025-08-07
437Qwen3 VL 4B (Reasoning)历史Alibaba7.05*21.3%12.2%91.8%4.6%0%开放2025-10-14
438Solar Pro 2 (Non-reasoning)历史Upstage7*0%16.1%93%3.7%0%闭源2025-07-09
439Mistral Small 3.2历史Mistral6.9629.2
裁判协议 v4.3.1
0
裁判协议 v4.3.1
0.3%0%28.6%20.3%14.8%76.8%0%4.3%0%开放2025-06-20
440Nova Pro历史Amazon6.96*21%16.9%77.7%3.2%0%闭源2024-12-03
441Command ACohere6.96*21.3%16.4%77.3%4%0%开放2025-03-13
442Qwen3.5 2BAlibaba6.94*165.7
裁判协议 v4.3.1
21%8.1%74.2%2.6%0%开放2026-03-02
443Llama 3.1 Nemotron 70BNVIDIA6.94*8.3%17.8%71.3%4.2%0%开放2024-10-15
444Llama 3.1 8B历史Meta6.93*-150.78
裁判协议 v4.3.1
18%8.5%43%5.3%0%开放2024-07-23
445Nemotron 3 Nano (Non-reasoning)NVIDIA6.85*-121.8
裁判协议 v4.3.1
10.7%11.4%91%4.6%0%开放2025-12-15
446NVIDIA Nemotron Nano 9B V2 (Non-reasoning)NVIDIA6.84*24%9.8%74.4%4.6%0%开放2025-08-18
447Qwen3 4B 2507 (Non-reasoning)历史Alibaba6.74*11.3%10.9%81.7%4.5%0%开放2025-08-06
448Qwen3 14B (Non-reasoning)历史Alibaba6.7*0%13.3%92.3%4.1%0%开放2025-04-28
449GPT-4历史OpenAI6.7*188.72
裁判协议 v4.3.1
21%69.4%闭源2023-03-14
450GLM-4.5V (Non-reasoning)历史Z AI6.7*0%18.2%90.2%3.5%0%开放2025-08-11
451Gemini 2.5 Flash-Lite (Non-reasoning)历史Google6.67*32%15.4%87.8%3.7%0%闭源2025-06-17
452Nova Lite历史Amazon6.67*19%9.8%57.6%4.3%0%闭源2024-12-03
453Hermes 4 70B (Non-reasoning)Nous Research6.66*4.7%18.7%82.7%3.6%0%开放2025-08-27
454Qwen3 30B (Non-reasoning)历史Alibaba6.62*0%12%88.5%4.6%0%开放2025-04-28
455Llama 3.1 70B历史Meta6.6*19.7%78.2%4.5%0%开放2024-07-23
456Granite 4.1 8B历史IBM6.57*13.3%12.3%87.1%3.8%0%开放2026-04-29
457Sarvam 30B (high)Sarvam6.56*0%12.6%96.3%7.5%0.3%开放2026-03-06
458Olmo 3.1 32B InstructAllen Institute for AI6.5*0%12%73.2%5%0%开放2026-01-13
459Gemma 4 E2B (Non-reasoning)Google6.48*16.3%8.2%74.4%4.7%0%开放2026-04-02
460Olmo 3 32B Think历史Allen Institute for AI6.47*0%14.9%85.7%6.4%0%开放2025-11-20
461Llama 4 ScoutMeta6.4559.71
裁判协议 v4.3.1
0
裁判协议 v4.3.1
0.2%0%21.3%27.7%15.2%79.4%0%3.8%0%开放2025-04-05
462Qwen3 14B历史Alibaba6.39187.72
裁判协议 v4.3.1
0
裁判协议 v4.3.1
0.2%0%30.7%0%15.3%75.8%1.2%4.5%0%开放2025-04-28
463Celeris-1Celeris6.35482.83
裁判协议 v4.3.1
171.51
裁判协议 v4.3.1
0.4%0%21.6%38%11%92.8%1.2%6.8%0%闭源2026-07-24
464Phi-4 MiniMicrosoft6.33*-171.32
裁判协议 v4.3.1
15.3%9.5%78%4.5%0%开放2024-02-26
465EXAONE 4.0 32B (Non-reasoning)LG AI Research6.33*10.6%82.1%5%0%开放2025-07-15
466Qwen3.5 2B (Non-reasoning)Alibaba6.19*-130.44
裁判协议 v4.3.1
14%7.4%97.7%5%0%开放2026-03-02
467Qwen3.5 0.8BAlibaba6.13*-152.17
裁判协议 v4.3.1
9%4.2%61.3%1.1%0%开放2026-03-02
468Jamba 1.7 LargeAI21 Labs6.07*19.3%20.6%95.1%3.7%0%开放2025-07-07
469Granite 4.0 H SmallIBM6.05*11.3%14.3%87.8%3.8%0%开放2025-09-22
470Ministral 3 14BMistral6.04432.06
裁判协议 v4.3.1
149.57
裁判协议 v4.3.1
0.6%0%23.8%26.3%13.6%92.5%1.2%4.6%0%开放2025-12-02
471Qwen3 Omni 30B A3BAlibaba6.01*0%14.3%97.6%4.6%0%开放2025-09-22
472Qwen3 8B (Non-reasoning)历史Alibaba5.99*0%11.1%95.6%1.9%0%开放2025-04-28
473LFM2 24B A2BLiquid AI5.95*0%6.5%69%4.2%0%开放2026-02-25
474Phi-4Microsoft5.92*0%14.1%81.2%3.8%0%开放2024-12-12
475Nova MicroAmazon5.88*13%10%65.1%4.6%0%闭源2024-12-03
476Granite 4.1 3B历史IBM5.86*73.87
裁判协议 v4.3.1
6%9.3%95.1%3.4%0%开放2026-04-29
477NVIDIA Nemotron Nano 12B v2 VL (Non-reasoning)NVIDIA5.82*20.7%11.9%95.5%4.3%0%开放2025-10-28
478MiniCPM-V 4.6 1.3BOpenBMB5.69*-134.2
裁判协议 v4.3.1
7%6.7%96.1%5.1%0%开放2026-05-11
479Jamba Reasoning 3BAI21 Labs5.67*6.3%6.7%69%3.8%0%开放2025-10-08
480Qwen3 VL 4B历史Alibaba5.66*14%10.9%97.4%3.6%0%开放2025-10-14
481Reka Flash 3Reka AI5.65*0%13.7%91.3%4.4%0%开放2025-03-10
482Olmo 3 7B ThinkAllen Institute for AI5.62*0%11.1%93.2%6%0%开放2025-11-20
483Claude 3 Haiku历史Anthropic5.58*27.7%17.6%80.5%4.1%0%闭源2024-03-04
484Ling-mini-2.0历史InclusionAI5.54*8.9%95.9%5.1%0%开放2025-09-09
485Ministral 3 8BMistral5.48404.6
裁判协议 v4.3.1
99.22
裁判协议 v4.3.1
0.5%0%20.7%25.7%13%94.2%0.8%4.3%0%开放2025-12-02
486Llama 3 70B历史Meta5.45*0%17.7%87.5%4.5%0%开放2024-04-18
487Llama 3.2 11B (Vision)Meta5.41*12.7%10.6%81.9%5.5%0%开放2024-09-25
488Qwen3.5 0.8B (Non-reasoning)Alibaba5.4*-133.2
裁判协议 v4.3.1
8%5%98.5%5.1%0%开放2026-03-02
489Sarvam M历史Sarvam5.31*0%15.2%90.6%3.1%0%开放2025-05-23
490Exaone 4.0 1.2BLG AI Research5.27*0%6.8%93%6%0%开放2025-07-15
491Olmo 3 7BAllen Institute for AI5.24*0%7.4%91.2%5.8%0%开放2025-11-20
492Exaone 4.0 1.2B (Non-reasoning)LG AI Research5.23*0%5%91.7%5.7%0%开放2025-07-15
493LFM2.5-1.2B-ThinkingLiquid AI5.22*0%7.4%94.4%6.2%0%开放2026-01-20
494Jamba 1.7 MiniAI21 Labs5.22*16%11.3%95.1%4.4%0%开放2025-07-07
495LFM2 2.6B历史Liquid AI5.21*0%5.4%62.9%5.5%0%开放2025-09-23
496LFM2.5-1.2B-InstructLiquid AI5.21*0%7%85.1%6.7%0%开放2026-01-05
497Granite 4.0 H 1BIBM5.19*7%5.2%81.7%5%0%开放2025-10-28
498Qwen3 1.7B历史Alibaba5.19*0%8.9%94.8%4.6%0%开放2025-04-28
499Qwen3 8B历史Alibaba5.16165.39
裁判协议 v4.3.1
0
裁判协议 v4.3.1
0.3%0%27.9%0%13.6%90.7%1.4%3.9%0%开放2025-04-28
500Gemma 3 270MGoogle5.12*0%0.9%30.4%3.6%0%开放2025-08-14
501Apertus 70B InstructSwiss AI Initiative5.11*0%11.6%82.8%5.5%0%开放2025-09-02
502Granite 4.0 MicroIBM5.11*7%9.3%95.8%5%0%开放2025-09-22
503Mistral 7B历史Mistral5.03*0%9%81.3%4.5%0%开放2023-09-27
504Granite 4.0 1B历史IBM5.01*6%6.2%93.5%4.8%0%开放2025-10-28
505Molmo2-8BAllen Institute for AI5*0%11.5%91.1%4.3%0%开放2025-12-11
506LFM2 8B A1B历史Liquid AI4.93*8.2%92.3%4.9%0%开放2025-10-07
507Granite 3.3 8B历史IBM4.92*9.7%96.2%4.2%0%开放2025-04-16
508Qwen3 1.7B (Non-reasoning)历史Alibaba4.86*0%6.9%97.8%5.3%0%开放2025-04-28
509Gemma 3 27B历史Google4.85-171.05
裁判协议 v4.3.1
0.2%0%23.3%7.3%13%92.1%0.2%4.4%0%开放2025-03-12
510Ministral 3 3BMistral4.84233.97
裁判协议 v4.3.1
16.01
裁判协议 v4.3.1
0.7%0%15.3%17%9%80.2%0%5.4%0%开放2025-12-02
511Qwen3 0.6B (Non-reasoning)历史Alibaba4.83*0%4.6%93.8%4.9%0%开放2025-04-28
512LFM2 1.2B历史Liquid AI4.83*0%4.9%81.4%5.6%0%开放2025-07-10
513Gemma 3 4B历史Google4.83*6.7%7.7%98.2%5.3%0%开放2025-03-12
514LFM2.5-VL-1.6BLiquid AI4.83*0%5.8%95.7%5.1%0%开放2026-01-05
515Qwen3 0.6B历史Alibaba4.83*0%6.3%90.9%5.6%0%开放2025-04-28
516Granite 4.0 350MIBM4.83*0%3.9%76.1%5.5%0%开放2025-10-28
517Llama 3.2 1B历史Meta4.83*6.7%7%66.3%5.5%0%开放2024-09-25
518Tiny Aya GlobalCohere4.83*0%6.1%96.2%5.2%0%开放2026-02-17
519Llama 3 8B历史Meta4.83*0%10.4%89.8%5.1%0%开放2024-04-18
520Apertus 8B InstructSwiss AI Initiative4.83*0%9.7%96.1%5%0%开放2025-09-02
521Granite 4.0 H 350MIBM4.83*0%3.8%88.1%6.4%0%开放2025-10-28
522Gemma 3n E4B历史Google4.83*0%8%97.1%4.5%0%开放2025-06-26
523Gemma 3n E2B历史Google4.83*0%6.8%93.4%4.2%0%开放2025-06-26
524Gemma 3 1B历史Google4.83*0%3.8%82.3%5.3%0%开放2025-03-13
525Gemma 3 12B历史Google3.75-169.68
裁判协议 v4.3.1
0.2%0%16.4%8.3%10.3%97.4%1%4.2%0%开放2025-03-12
526K2 Horizon 0.9BInstitute of Foundation Models3.05-173.76
裁判协议 v4.3.1
0
裁判协议 v4.3.1
0.2%0%6.9%6.3%6.9%85.7%0.8%5.4%0%开放2026-09-03
527Mi:dm K 2.5 Pro Preview历史Korea Telecom14.7%16.4%94.7%9.1%0%闭源2025-12-11
528Cogito v2.1Deep Cogito22.7%30.2%80.2%12%0%开放2025-11-18

v4.1 历史参考与版本边界

旧版包含 9 项评测:GDPval-AA v2、τ³-Banking、Terminal-Bench v2.1、SciCode、AA-LCR、AA-Omniscience、Humanity's Last Exam、GPQA Diamond、CritPt。v4.3 引入新的组成项、任务版本与权重,不能把换版本后的分数差直接解释成模型进步或退步。

查看当前权重与旧版方法对照 →

实时 Elo 的裁判协议与可比边界

AA-Briefcase 与 GDPval-AA v2 的当前成对比较使用裁判协议 v4.3.1,原评测名称与任务版本不变,Intelligence Index 仍为 v4.3。未区分裁判协议的旧 Elo 保留原始观测,仅作历史参考,不与当前协议混排。

指数中的 Elo 贡献在模型加入时冻结;实时榜单分数不能用于重算指数。来源未给出逐模型重跑日期,不能把裁判更新或跨协议分数差解释成新的模型运行或能力变化。

查看裁判、核验时间与完整方法边界 →查看 AA 原始方法 ↗

FAQ

智能指数拆分 常见问题

回答只使用来源页面可以明确核对的指标定义。

AA Intelligence Index v4.3 包含哪些评测?

包含 AA-Briefcase、GDPval-AA v2、AutomationBench-AA、Terminal-Bench v4.0、SciCode、AA-Omniscience、GDP.pdf、AA-LCR v1.1、Humanity's Last Exam 和 CritPt,共 10 项;Omniscience 的准确率与非幻觉率分别计入权重。

v4.3 和 v4.1 的分数可以直接比较吗?

不可以直接把跨版本分数差解释成能力变化。两版的任务集、组成评测和权重不同;本站按版本展示,旧观测保留 v4.1 身份,新观测不覆盖旧历史。

来源估算的 Intelligence Index 代表什么?

估算是 Artificial Analysis 对该结果的标记,不等同于全部组成评测均已完成。本站保留估算标记与原始分数,缺失组成字段显示为未知,不自行补算。

裁判协议 v4.3.1 是否意味着 Intelligence Index 也换成 v4.3.1?

不是。该方法补丁更新 AA-Briefcase 与 GDPval-AA v2 的实时 Elo 成对裁判,Intelligence Index 仍标为 v4.3;指数中的 Elo 贡献在模型加入时冻结,不能用实时 Elo 重算。旧观测未区分裁判协议时仅作历史参考。

这个评测关注什么

按方法版本拆解知识、推理、编程、工具使用和长上下文表现,不混排不同版本。

阅读时注意

评测结果受任务集合、评分方法和模型版本影响。本站保留原始口径,不用它生成新的综合排名。

适用场景

用于回答一个明确的问题:模型在这一类任务上的相对表现如何,而不是“哪个模型绝对最好”。