JPMorgan has issued a research report reiterating its "Overweight" rating on Knowledge Atlas (02513) and raising its target price from HK$1,400 to HK$1,800. This new target is derived from a discounted valuation based on 30 times the projected 2030 price-to-earnings ratio.

The bank has increased its revenue forecasts for the company's artificial intelligence segment, citing the enhanced capabilities and integrated pricing of the GLM-5.2 model. The firm now holds a more optimistic view of Knowledge Atlas's prospects following the release of this model and its updated pricing structure.

The report notes that compared to GLM-5.1, the GLM-5.2 version has eliminated the lower billing tiers, applying a single, higher rate across all usage, resulting in an overall 13% increase in API pricing. As GLM-5.2 remains part of the GLM-5 model family, which features 744 billion total parameters and 40 billion active parameters, its performance improvements are primarily driven by reinforcement learning and post-training optimization. This allows the company to achieve higher effective pricing while maintaining a largely stable underlying model cost structure.

JPMorgan observes a bifurcation trend in the large language model pricing market. Pricing for mature, general-purpose products continues to decline due to widespread capability availability and improving inference costs, as exemplified by models like DeepSeek, which persistently pressures market clearing prices for standard, price-sensitive workloads. Conversely, newly unlocked cutting-edge capabilities that improve task completion, reduce retries, and enable high-value applications like coding, intelligent agents, and enterprise workflow automation can still command a premium.

Knowledge Atlas's GLM-5.2 serves as a useful test case for pricing power. If demand for its APIs and workflows remains resilient following the effective price increase, it would further confirm the company's ability to translate model iterations into high-quality revenue growth with incremental profitability.

The bank highlighted several near-term factors requiring attention, including the demand resilience for GLM-5.2, the progress of model updates from peers such as Kimi K3 and DeepSeek V4.1, and the anticipated launch of Knowledge Atlas's GLM-5.5 model expected in August.