GO
| HSI1 | 23,446.34 | +110.06 | 257.13B |
| HSCEI1 | 7,773.67 | +14.31 | 71.88B |
| Back Zoom + Zoom - Block Traded | |
|
2026-06-22 14:05:32 CLSA published a report stating that since the official release of KNOWLEDGE ATLAS (02513.HK)'s GLM-5.2 API on June 17, substantial evidence has emerged showing that Zhipu AI has advanced to global state-of-the-art (SOTA) levels in code generation, with performance comparable to Opus 4.6 to 4.8, despite having a much smaller model size than its peers. The global community is also optimistic that China could achieve Fable-level AI capabilities in 1Q27, or even earlier. CLSA views this as a landmark moment marking formal global recognition of China癒礎s post-training capabilities, and strong proof of Zhipu AI癒礎s data flywheel effect. The broker believes this reflects a major victory for Zhipu in reinforcement learning, describing it as a genuine "GLM Moment," coming 16 months after DeepSeek癒礎s breakthrough in pre-training and infrastructure. The report noted that GLM-5.2 has significantly enhanced the appeal of Zhipu AI癒礎s API services, but believes that near-term ARR (annual recurring revenue) growth will still largely depend on available computing power resources. Official pricing remains in line with GLM-5.1 for inputs above 32k, at approximately RMB8 per million tokens for input and RMB28 per million tokens for output. However, if enterprises heavily utilize one million-token contexts, such memory-intensive applications could lead to a significant increase in per-token inference costs. Therefore, more data is needed to assess margin trends. The broker assigned a Hold rating to Zhipu. (ha/da)~ AASTOCKS Financial News Website: www.aastocks.com This article was automatically translated by AI, the Chinese version should be considered the authoritative version. AASTOCKS.com Limited does not guarantee its accuracy or completeness and accepts no liability for any damages or losses arising from the use of this translation. | |