AI 文档理解模型排行榜

查看 Arena 文档任务榜的评分。

评测来源 Arena · 文档原榜发布 查看原榜评分方法

Arena Score

46 个模型
Arena · 文档 · Arena Score
排名模型 / 开发机构Arena Score投票95% CI许可详情
01Claude Fable 5 (High)Anthropic1499Arena Score投票1260095% CI1493–1506许可Proprietary投票1260095% CI1493–1506许可Proprietary来源
02Claude Fable 5.1 (Max)Anthropic1499Arena Score投票147695% CI1484–1514许可Proprietary投票147695% CI1484–1514许可Proprietary来源
03Claude Opus 4.6 (High)Anthropic1496Arena Score投票2969595% CI1490–1503许可Proprietary投票2969595% CI1490–1503许可Proprietary来源
04Claude Opus 4.6Anthropic1495Arena Score投票4777895% CI1490–1501许可Proprietary投票4777895% CI1490–1501许可Proprietary来源
05Claude Opus 4.7Anthropic1495Arena Score投票2383095% CI1488–1501许可Proprietary投票2383095% CI1488–1501许可Proprietary来源
06Claude Opus 4.7 (High)Anthropic1494Arena Score投票2356795% CI1488–1501许可Proprietary投票2356795% CI1488–1501许可Proprietary来源
07GPT-5.6 Sol (xHigh)OpenAI1486Arena Score投票978295% CI1479–1493许可Proprietary投票978295% CI1479–1493许可Proprietary来源
08Claude Opus 5 (High)Anthropic1486Arena Score投票1665995% CI1480–1491许可Proprietary投票1665995% CI1480–1491许可Proprietary来源
09Claude Opus 4.8 (High)Anthropic1486Arena Score投票1218095% CI1479–1492许可Proprietary投票1218095% CI1479–1492许可Proprietary来源
10GPT-5.5 (High)OpenAI1482Arena Score投票2777495% CI1477–1487许可Proprietary投票2777495% CI1477–1487许可Proprietary来源
11GPT-5.5OpenAI1480Arena Score投票2818295% CI1474–1485许可Proprietary投票2818295% CI1474–1485许可Proprietary来源
12Muse Spark 1.1Meta1479Arena Score投票916695% CI1472–1486许可Proprietary投票916695% CI1472–1486许可Proprietary来源
13Claude Sonnet 4.6Anthropic1478Arena Score投票6151195% CI1472–1484许可Proprietary投票6151195% CI1472–1484许可Proprietary来源
14GPT-6 Astra (Max)OpenAI1477Arena Score投票584295% CI1469–1486许可Proprietary投票584295% CI1469–1486许可Proprietary来源
15Claude Opus 4.8Anthropic1476Arena Score投票1493895% CI1469–1482许可Proprietary投票1493895% CI1469–1482许可Proprietary来源
16Muse Spark 1.3 (Max)Meta1474Arena Score投票733295% CI1467–1482许可Proprietary投票733295% CI1467–1482许可Proprietary来源
17GPT-5.6 Terra (xHigh)OpenAI1474Arena Score投票1176995% CI1468–1480许可Proprietary投票1176995% CI1468–1480许可Proprietary来源
18GPT-5.4OpenAI1471Arena Score投票3542595% CI1465–1477许可Proprietary投票3542595% CI1465–1477许可Proprietary来源
19Claude Sonnet 5 (High)Anthropic1469Arena Score投票1127095% CI1463–1476许可Proprietary投票1127095% CI1463–1476许可Proprietary来源
20Gemini 3.8 Flash (High)Google1469Arena Score投票708595% CI1461–1477许可Proprietary投票708595% CI1461–1477许可Proprietary来源
21Gemini 3.7 Flash (High)Google1468Arena Score投票783095% CI1460–1476许可Proprietary投票783095% CI1460–1476许可Proprietary来源
22Claude Opus 4.5 (20251101)Anthropic1468Arena Score投票836495% CI1458–1478许可Proprietary投票836495% CI1458–1478许可Proprietary来源
23Muse SparkMeta1466Arena Score投票113095% CI1448–1484许可Proprietary投票113095% CI1448–1484许可Proprietary来源
24Grok 4.5xAI1466Arena Score投票758995% CI1459–1474许可Proprietary投票758995% CI1459–1474许可Proprietary来源
25GPT-5.6 Luna (xHigh)OpenAI1460Arena Score投票902395% CI1453–1467许可Proprietary投票902395% CI1453–1467许可Proprietary来源
26GPT-5.5 InstantOpenAI1459Arena Score投票888495% CI1451–1468许可Proprietary投票888495% CI1451–1468许可Proprietary来源
27Gemini 3.1 Pro PreviewGoogle1459Arena Score投票5263495% CI1454–1464许可Proprietary投票5263495% CI1454–1464许可Proprietary来源
28Gemini 3.5 Flash (Medium)Google1458Arena Score投票690895% CI1451–1466许可Proprietary投票690895% CI1451–1466许可Proprietary来源
29Gemini 3.5 Flash (High)Google1455Arena Score投票429295% CI1445–1464许可Proprietary投票429295% CI1445–1464许可Proprietary来源
30Gemini 3.6 Flash (High)Google1454Arena Score投票307595% CI1443–1464许可Proprietary投票307595% CI1443–1464许可Proprietary来源
31Grok 4.6 (High)xAI1453Arena Score投票779395% CI1445–1460许可Proprietary投票779395% CI1445–1460许可Proprietary来源
32Gemma 4 31BGoogle1451Arena Score投票1556895% CI1444–1458许可Apache 2.0投票1556895% CI1444–1458许可Apache 2.0来源
33Gemini 3 ProGoogle1451Arena Score投票1141695% CI1442–1460许可Proprietary投票1141695% CI1442–1460许可Proprietary来源
34Kimi K2.6Moonshot AI1451Arena Score投票1190995% CI1443–1458许可Modified MIT投票1190995% CI1443–1458许可Modified MIT来源
35Claude Sonnet 4.5 (20250929)Anthropic1448Arena Score投票3316695% CI1442–1454许可Proprietary投票3316695% CI1442–1454许可Proprietary来源
36Qwen3.7-PlusAlibaba1446Arena Score投票751195% CI1439–1453许可Proprietary投票751195% CI1439–1453许可Proprietary来源
37Grok 4.20 Beta (Reasoning, 0309)xAI1439Arena Score投票2211195% CI1433–1446许可Proprietary投票2211195% CI1433–1446许可Proprietary来源
38Gemini 3 FlashGoogle1439Arena Score投票759395% CI1430–1449许可Proprietary投票759395% CI1430–1449许可Proprietary来源
39Kimi K2.5 (Thinking)Moonshot AI1437Arena Score投票2277095% CI1430–1444许可Modified MIT投票2277095% CI1430–1444许可Modified MIT来源
40MiniMax-M3MiniMax1435Arena Score投票666995% CI1427–1444许可MiniMax Community License投票666995% CI1427–1444许可MiniMax Community License来源
41Gemini 2.5 ProGoogle1432Arena Score投票2681895% CI1426–1438许可Proprietary投票2681895% CI1426–1438许可Proprietary来源
42GPT-5.2OpenAI1425Arena Score投票2991895% CI1419–1431许可Proprietary投票2991895% CI1419–1431许可Proprietary来源
43GPT-5.2 (High)OpenAI1421Arena Score投票758495% CI1412–1431许可Proprietary投票758495% CI1412–1431许可Proprietary来源
44Claude Haiku 4.5 (20251001)Anthropic1419Arena Score投票3977195% CI1413–1424许可Proprietary投票3977195% CI1413–1424许可Proprietary来源
45GPT-5.1OpenAI1416Arena Score投票877395% CI1407–1425许可Proprietary投票877395% CI1407–1425许可Proprietary来源
46GLM-5V-TurboZ.ai1407Arena Score投票926395% CI1399–1414许可Proprietary投票926395% CI1399–1414许可Proprietary来源

数据来自 Arena 官方公开数据集(CC BY 4.0);保留原榜名次与评分。 官方数据集

未公布的数据记为 —