CICC: Maintains "outperform industry" rating on KNOWLEDGE ATLAS (02513), raises the target price to HK$1250.

CICC released a research report stating that it maintains its profit forecast for KNOWLEDGE ATLAS (02513) unchanged. Due to the company's fast iterative model capabilities exceeding market and industry expectations, CICC is more optimistic about the company's long-term API accelerated release, and has raised the target price by 39% to HK$1250 (corresponding to 67x P/S in 2027). This represents a 14% upside potential from the current stock price. CICC maintains its "outperform industry" rating, with the company currently trading at 58x P/S in 2027. Key points from CICC include: Company Updates On June 13, the company's latest model GLM-5.2 was opened to all users of GLM Coding Plan, covering Lite/Pro/Max/Team versions. Additionally, the company announced that the GLM-5.2 API will be officially launched and open-sourced in the third week of June, following the MIT license. GLM-5.2 is the company's latest and most intelligent open source model, focusing on coding direction and improving the long-context capability to 1M Tokens. It has received excellent feedback from many developers and is the first domestic model with real capabilities comparable to Claude-Opus 4.6, ranking first in domestic model programming capabilities and comparable to top overseas models. Continued focus on coding field, significant improvement in front-end and back-end programming capabilities According to CICC and many developers' evaluations, GLM-5.2 has shown significant progress in overall programming capabilities compared to GLM-5.1. It has excelled in backend evaluations and can now match Claude Opus 4.6. Front-end tasks have also shown noticeable improvements in aesthetics, precision, and controllability. GLM-5.2 remains the top domestic model in terms of comprehensive programming capabilities and is gradually distancing itself from other domestic models. Long-context capability improved with 1M Tokens CICC believes that GLM-5.2 has made significant optimizations at the model infrastructure level, increasing the context length from 200K to 1M Tokens. This improvement bridges the capability gap in long-context tasks that GLM-5.1 had, significantly enhancing its usability in real complex programming scenarios and improving its ability to handle complex multi-turn Agent tasks. Actively addressing computational power bottlenecks to enhance user experience and supply efficiency Due to the company's outstanding model capabilities, there is a demand-supply imbalance causing constraints on computational power. Users may experience longer processing times with GLM-5.2, primarily due to high concurrency and limited computational power. CICC believes that the company will actively address these computational power issues to alleviate supply side bottlenecks. Risk Warning: Model iteration not as expected; user feedback lower than expected.