SenseTime Raccoon

What it is SenseTime Raccoon (商汤小浣熊, also called 小浣熊) is a set of AI products developed by SenseTime for work tasks. Its branches include Office Raccoon, which focuses on office and data work, and Code Raccoon, which focuses on programming ^[1][2]. Together they form one product family under the same name.

Origins SenseTime presented Office Raccoon together with SenseNova 4.0 in February 2024. It was described as a no-code data-analysis tool that turns requests written in natural language into analysis and charts ^[1]. No-code means that a user does not need to write programs to get results, and natural language means ordinary written instructions rather than a specialised query language.

Main features In July 2024 Code Raccoon gained responses five times faster and 10% higher coding precision. In the same update, Office Raccoon added a consumer web page and a WeChat mini-app ^[2]. A WeChat mini-app is a lightweight program that runs inside the WeChat messaging app, without a separate installation.

The product also came in industry editions. The education edition served more than 500 institutions, and there was also a finance edition ^[3].

Its place in development In July 2025 Raccoon received an agent upgrade powered by SenseNova V6.5. SenseTime said its user base had exceeded 10 million ^[3]. The company also claimed that the upgraded Raccoon was on par with Claude 4 Opus in data analysis and outperformed OpenAI o3 ^[3]. These are SenseTime's own claims and were not presented as independent test results.

In the first half of 2026 Raccoon served Lenovo, Ping An Technology, China's three major telecom operators, JD.com and Kylin Software. Total users grew more than fivefold compared with the same period a year earlier ^[4].

For readers less familiar with the terms, an AI agent is a system that can plan and carry out a series of actions with tools such as data sources, spreadsheets or software interfaces, rather than only generating text. Data analysis tools that accept plain-language requests are one common use of this approach, because the user describes the result wanted and the system works out the steps to produce it.