Public Notice of Algorithm and Model Filing Information for Kimi

Last Updated: August 24, 2026

Beijing Moonshot Technology Co., Ltd. ("Moonshot," "we," "us," or "our") operates Kimi products and services, including but not limited to the products and services we provide to you through our website (kimi.com), applications (the Kimi APP and its PC version; the Kimi Claw APP), mini-programs (Kimi Smart Assistant), and innovative forms that emerge with technological development (hereinafter "products and/or services"). To protect users' right to know and to fulfill relevant legal requirements, we hereby publish the filing information of the models and algorithms used in connection with Kimi as follows.

The model for which we have completed the filing is as follows:

Model Name: Moonshot (Kimi Large Model)

Filing Number: Beijing-MoonShot-20231016

Model Application Scenarios: Moonshot (Kimi Large Model) is mainly used in products and/or services such as Kimi Smart Assistant (website, APP, and WeChat mini-program) for scenarios including intelligent Q&A, conversational interaction, text generation, information search, and content summarization, and can generate corresponding text responses based on the questions you enter and the content you upload.

Link to Public Information: CAC Announcement Page

We have completed the filing for two algorithms in total. The filing information is published as follows:

Basic Principles of the Algorithm: This algorithm is a pre-trained language model independently developed based on the Transformer architecture. It consists of multiple stacked Transformer encoders, each containing a multi-head self-attention mechanism and a feed-forward neural network. It is capable of processing long text sequences and generating natural text responses based on context.

Operating Mechanism of the Algorithm: After a user inputs natural language data in text format or uploads text content, the algorithm first reviews the input data for illegal and harmful information. If the review is not passed, the user will be informed that the service cannot be provided. If the review is passed, the data is fed into the algorithm, which performs semantic and contextual correlation analysis and generates a response in natural language. The output data is then reviewed again for illegal and harmful information, and the response is provided to the user only after passing this review.

Algorithm Application Scenarios: Applied in the knowledge-base Q&A scenarios of Kimi Smart Assistant on the website, the APP (Android and iOS), and the WeChat mini-program.

Purpose and Intent of the Algorithm: Dedicated to improving the text processing efficiency of enterprises and individuals, and to providing users with Q&A and dialogue generation services in general scenarios through text deep synthesis technology, thereby achieving purposes such as conversational interaction, answering questions, and assisting in content creation.

Filing Number: Wang Xin Suan Bei No. 110108896786101240015

Link to Public Information: Internet Information Service Algorithm Filing System

Basic Principles of the Algorithm: This algorithm is a pre-trained language model independently developed based on the Transformer architecture. It consists of multiple stacked Transformer encoders, each containing a multi-head self-attention mechanism and a feed-forward neural network. It is capable of processing long text sequences and generating natural text responses based on context.

Operating Mechanism of the Algorithm: After a user inputs natural language data in text format or uploads text content, the algorithm first reviews the input data for illegal and harmful information. If the review is not passed, the user will be informed that the service cannot be provided. If the review is passed, the data is fed into the algorithm, which performs semantic and contextual correlation analysis and generates a response in natural language. The output data is then reviewed again for illegal and harmful information, and the response is provided to the user only after passing this review.

Algorithm Application Scenarios: Applied to the Kimi Open Platform.

Purpose and Intent of the Algorithm: Serving developers with needs for text generation and intelligent dialogue development, dedicated to improving the text processing efficiency of enterprises and individuals, and providing users with Q&A and dialogue generation services in general scenarios through API calls to the algorithm model, thereby achieving purposes such as conversational interaction, answering questions, and assisting in content creation.

Filing Number: Wang Xin Suan Bei No. 110108896786101240023

Link to Public Information: Internet Information Service Algorithm Filing System

The filing information listed in this notice shall be subject to the information published by the competent authorities, such as the Internet Information Service Algorithm Filing System. In the event of any change to the relevant services, algorithms, model names, or filing numbers, we will update this notice in a timely manner in accordance with legal and regulatory requirements and actual circumstances.