The Chinese evaluation group SuperCLUE published the results of its SuperCLUE-Office benchmark, which assesses the performance of office agents at the end of October 2026. According to data published on the SuperCLUE website and relayed by IT Home, ByteDance's Doubao Work product secured the top spot with an overall score of 85.33.
Kimi Work from Moonshot AI took second place with a result of 84.57, while WorkBuddy from Tencent ranked third with a score of 83.99. It is important to note that this ranking is an industry benchmark from SuperCLUE, and Pandaily publishes the organizer's assessments, not independent retesting.
The SuperCLUE-Office test checks the ability of office agents to perform real-world tasks in programs such as Word, PowerPoint, Excel, and charts. The October round covered three areas: document processing, material composition, and data analysis, within 13 secondary tasks. Five competency areas were evaluated: file and environment operations, multi-file planning, constraint adherence, content quality, and security and permission boundaries.
Eight Chinese office agents participated in the competition: Doubao Work, Kimi Work, WorkBuddy, Baidu Dazi, QwenWork from Alibaba, Raccoon Work from SenseTime, Spark Work from iFLYTEK, and Nano Work from 360.
The leaders in the ranking were quite close, as the top four shared a difference of only 2.65 points. Baidu Dazi scored 82.68, QwenWork scored 78.38, Raccoon Work scored 69.20, and iFLYTEK Spark Work scored 69.12, while Nano Work received 53.13. SuperCLUE noted that the overall gap between the highest and lowest scores was 32.20 points, indicating a clear performance gradient below the leading group.
The leaders demonstrated differences across categories. Doubao Work led the document processing category with a score of 87.12, surpassing the next participant by almost five points; all eight products scored above 72 in this area, with Baidu Dazi and WorkBuddy trailing behind. Kimi Work led in data analysis with a score of 88.00, followed by WorkBuddy (86.13) and Doubao Work (81.50). SuperCLUE indicated that this category had the largest spread, with a difference of 72 points between the best and worst results and an average of 71.72. In the material composition category, the difference was smaller but uneven further down the table, reaching a maximum gap of 47.75 points with an average of 73.62.
SuperCLUE concluded that office agents are gradually integrating into daily workflows, but they remain heterogeneous when performing more complex tasks related to analysis and document creation, which was noted as the next focus of competition. Since the results are based on a set of tasks and tools from one organizer, readers should view them as a public ranking, not as a universal classification of all office agents.



