Domestic lobster Claw showdown: Baidu first, Xiaomi second

Kuai Technology reported on August 7 that the evaluation agency SuperCLUE released the second complete list of XClaw lobster agents. A total of 10 mainstream domestic office AIs were included this time. The average scores of all products in the three tests exceeded 91 points. The overall level was greatly improved. Baidu Wenxin Assistant took the first place in the total score, and Xiaomi MiMo Claw ranked second. The entire set of assessments is completely consistent with real work in the workplace and is divided into five core competency sections, namely data processing, content creation, memory ability, code development, and research analysis. The test will not only run the results once. Each task will be repeated three times independently, and the average value will be used as the final score to avoid errors caused by single fluctuations. Judging from the overall list, the average score of the 10 products reached 94.53 points. The difference between the first and last place was only 6.31 points. The scores of the top four products all exceeded 96 points. The competition was very fierce. Baidu Wenxin Assistant ranks first with a total score of 97.62 points. The second place Xiaomi MiMoClaw and the third place Tencent WorkBuddy are only 0.13 points apart. The small difference can show that the comprehensive strength of the head products is very close. Comparison by dimension shows clear differentiation between strengths and weaknesses. All products in the two major sectors of content creation and memory score are close to full marks. Basic office needs such as writing copywriting and retaining context for long conversations can be basically completed stably by each company. Data processing and industry survey scores came second. Only the code development dimension opened up a huge gap. The industry average score was only 86.99 points, and the difference between the highest score and the lowest score was 17.59 points. Among them, Xiaomi MiMo Claw performed best in the code development project, with a score of 97.22, leading other lobsters in terms of fault. Stability is also a key indicator in this evaluation. The smaller the fluctuations in the three test scores, the more stable the performance in daily use. WorkBuddy and MaxClaw have the lowest fluctuation values, and the output results are highly consistent every time. Some products with lower total scores have huge differences in multiple tests. The output content of the same command is uneven each time, making it easy to make mistakes in daily office work. Now domestic office AI has got rid of the competition of basic functions and entered into details, stability and professional tasks.