KNOWLEDGE ATLAS (02513.HK) +17.000 (+1.167%) Short selling $92.75M; Ratio 2.042% , after surging nearly 33% yesterday, extended its rally this morning and is now up 9.27% at HKD1,592. Following JPM’s earlier 47% increase in its TP to HKD1,400, CICC released a report yesterday (15th), expressing optimism that the company’s leading position in coding capabilities will be further consolidated. CICC raised its TP by 39% to HKD1,250 and maintained an Outperform (industry) rating.The report noted that on June 13, the company announced that its latest model, GLM-5.2, is now fully available to all users under the GLM Coding Plan, covering Lite/Pro/Max/Team editions. Meanwhile, the GLM-5.2 API will officially launch and be open-sourced in the third week of June under the MIT license. As the company’s newest and most advanced open-source model, GLM-5.2 continues to focus on coding capabilities and has expanded its long-context capacity to 1M Tokens. Its long-horizon task performance has improved significantly and has received strong feedback from developer evaluations. It is the first domestic model whose real-world capabilities have reached the level of Claude-Opus 4.6, ranking first among domestic models in coding performance and benchmarking against leading overseas models.Related News Macquarie: KNOWLEDGE ATLAS (02513.HK) Launch of GLM-5.2 Positive, Enhances Pricing PowerAccording to the report, based on assessments by the brokerage and numerous developers, GLM-5.2 has demonstrated notable overall improvements in coding capability compared with GLM-5.1. In backend domains where it already had strengths, multiple evaluation tasks now benchmark against Claude Opus 4.6. For frontend tasks, aesthetics, refinement and controllability have also improved significantly. Its comprehensive coding capability continues to rank first among domestic models and it is the first domestic model to genuinely reach Opus 4.6-level performance, gradually widening the gap with other local peers.With context length increased to 1M Tokens, long-horizon task capability has improved markedly. The brokerage believes GLM-5.2 incorporates substantial optimizations at the model architecture level, expanding context length from 200K to 1M Tokens. This addresses GLM-5.1’s previous shortcomings in long-horizon tasks, significantly enhancing usability in real-world complex programming scenarios, while its ability to complete complex multi-round agent tasks has also improved materially. (ha/u)(HK stocks quote is delayed for at least 15 mins.Short Selling Data as at 2026-06-16 16:25.)
AASTOCKS Financial News