JP Morgan now believes that in China's AI model layer competition, internally driven capability improvements offer stronger investment value than passive environmental enhancements. Z.AI has secured a more advantageous competitive starting point through the inherent capability leap of GLM-5.3, while MiniMax's valuation revaluation depends on whether M3.1 and H3 can deliver convincing results in their respective dimensions.

According to a report from the Chasing Flow Trading Desk, JP Morgan's August 16 research note highlights two key catalysts—the release of Z.AI's GLM-5.3 and DeepSeek's API price hike effective August 17—that are reshaping the industry's competitive landscape. The core conclusion is clear: in the rapidly evolving AI market, models with "frontier intelligence" pricing power are more valuable investments than those relying solely on low prices.

Why Capability-Driven Models Are Favored

JP Morgan argues that the current preference for capability is justified because the intelligence frontier is still advancing quickly. Z.AI's enhanced competitiveness stems from internal drivers, making its moat more durable. Meanwhile, MiniMax must prove itself through the upcoming M3.1, establishing a competitive advantage on some dimension of the Pareto curve. The Hailuo H3 offers multi-modal option value, but independent vendors face uncertainty in capturing value against integrated platforms like ByteDance and Kuaishou.

Based on this logic, JP Morgan has made the following key investment rating and target price adjustments:

Z.AI: Maintained an "Overweight" rating, with the December 2026 target price significantly raised from HK$1,600 to HK$1,800. Z.AI's internally driven technological advancement (strengthened post-training) has enabled the capability leap of GLM-5.3, solidifying its moat.

MiniMax: Maintained a "Neutral" rating, with the December 2026 target price raised from HK$160 to HK$260. The positive factors are mainly the breathing room provided by competitor DeepSeek's price hike and the multi-modal option value from Hailuo H3. However, its core large model capabilities still need validation from the upcoming M3.1.

Simultaneously, JP Morgan has also raised its earnings forecasts for both companies: Z.AI's 2026-27 revenue forecast is up 6-9%, with a 6-10% overall increase for 2026-30. MiniMax's 2026 revenue forecast remains unchanged, while its 2027-30 revenue forecast is raised by 11-21%.

Competition Framework: The Pareto Frontier Defines the Two-Track Race of Capability and Cost

JP Morgan introduces the "Pareto Frontier" as the core framework for evaluating China's AI model competition. It is defined as: when no competitor offers stronger capabilities at the same or lower price, or comparable capabilities at a lower price, that model is on the Pareto Frontier.

Models on the frontier can build two types of attractive business models:

JP Morgan currently favors the capability side for three reasons. First, model intelligence is still rapidly improving, and capability-leading models are relatively less affected by price changes from weaker substitutes. Second, each level of capability leap can unlock new demand—for example, in programming, we have evolved from auto-completion to warehouse-level development and long-cycle software development tasks. Third, as intelligence matures, the cost-performance track will become more competitive, and a sustainable cost leadership position requires a structural efficiency advantage, not aggressive pricing willingness.

On the cost side, JP Morgan primarily uses total token pricing as a metric (assuming an input-to-output ratio of 10:1 and a cache hit rate of 90%). On the capability side, it references benchmark tests and actual product performance.

DeepSeek's Price Hike: Industry Cost Anchor Loosens, but Structural Advantage Remains

DeepSeek's API price adjustment, effective August 17, is a significant background event in JP Morgan's report. The pricing data shows substantial adjustments:

V4 Pro (peak hours): Input price raised from 3.00 yuan/million tokens to 9.00 yuan, output from 6.00 yuan to 27.00 yuan.

V4 Flash (peak hours): Input raised from 1.00 yuan to 3.00 yuan, output from 2.00 yuan to 9.00 yuan.

V4 Flash (off-peak): Input slightly increased from 1.00 yuan to 1.50 yuan, output from 2.00 yuan to 4.50 yuan.

JP Morgan interprets this price hike as indicating that DeepSeek's previous pricing contained significant monetization room, and the increase demonstrates its ample pricing flexibility. Despite this, DeepSeek still maintains strong price competitiveness among high-capability models, and its status as the industry's cost benchmark remains unchanged.

The impact of this price hike on the industry is twofold. In the short term, it provides breathing room for vendors near the cost-performance end, especially MiniMax, narrowing their cost disadvantage. However, in the long term, it reinforces JP Morgan's core view: a sustainable cost leadership position requires underlying efficiency advantages that persist after price adjustments. DeepSeek's system-level efficiencies, such as its MoE architecture, attention design, and KV cache optimization, remain its moat.

Z.AI: GLM-5.3 Internally Drives Capability Leap, Maintain 'Overweight'

According to Z.AI's disclosures, GLM-5.3 uses the same base model as GLM-5.2, with improvements in programming and agent capabilities primarily stemming from reinforced post-training.

JP Morgan believes this technical path has significant competitive implications: it shows that substantial capability improvements can be achieved through post-training alone, without a new round of large-scale pre-training. This increases the weight of data quality, reinforcement learning, evaluation infrastructure, and engineering execution in differentiating competition.

With the API pricing system largely unchanged (GLM-5.2 pricing remains at 8.00 yuan for input, 2.00 yuan for cache hit input, and 28.00 yuan for output per million tokens), the stronger capabilities are expected to improve adoption and retention rates.

JP Morgan emphasizes that Z.AI's enhanced competitiveness is internally driven, which is the core reason it is more attractive than MiniMax. Starting from a point closer to the intelligence frontier also gives Z.AI strategic flexibility: through future inference optimization, it can improve cost-performance while maintaining model capability.

The sustainability of the investment logic requires continuous model iteration, not just the single release of GLM-5.3. Z.AI's competitive position still faces challenges from subsequent products from Kimi, DeepSeek, and other frontier peers.

JP Morgan has raised its 2026-30 revenue forecast for Z.AI by 6-10%, with specific adjustments as follows:

The target price of HK$1,800 is based on 20 times the expected 2030 P/E ratio, discounted to December 2026 using a 15% weighted average cost of capital. The 20x P/E ratio represents a valuation premium over leading domestic internet companies, primarily reflecting the company's expected revenue compound annual growth rate of over 100% from 2026 to 2030.

MiniMax: External Tailwinds Provide Breathing Room, M3.1 and H3 Are Two Upward Paths

JP Morgan notes that MiniMax's current M3 model has not established a clear advantage in either capability or cost-performance, putting it in a difficult position. It faces pressure from stronger models (like Kimi K3, GLM-5.3) on one side and from DeepSeek's long-term aggressive pricing strategy on the other.

DeepSeek's price hike narrows MiniMax's cost disadvantage in substitutable workloads, supporting higher retention rates, workload share, and pricing assumptions. However, JP Morgan clearly states that this benefit stems from a competitor's decision and could reverse with changes in pricing and models.

JP Morgan positions M3.1 as MiniMax's most important company-level catalyst. The evaluation criteria are clear: either significantly improve model capability or create outstanding cost-performance, establishing a more competitive position on any dimension of the Pareto curve. If M3.1 is only a modest improvement and the model remains within the Pareto frontier, the impact on the long-term view is limited.

Hailuo H3 adds a second potential growth engine for MiniMax, with initial positive market feedback enhancing its multi-modal product portfolio. As AI-generated image and video technology penetrates into advertising, short-form video, gaming, and e-commerce industries, the demand outlook is broad.

However, JP Morgan is cautious about MiniMax's value capture capability as an independent supplier. The core challenge is the competitive landscape: integrated platforms like ByteDance and Kuaishou can capture value across multiple stages—model/API revenue, content creation, distribution, advertising, and user interaction—and have existing creator and advertiser ecosystems providing distribution and data advantages. In contrast, MiniMax must more directly generate returns through its own products and model services.

JP Morgan has raised its 2027-30 revenue forecast for MiniMax by 11-21%, with specific adjustments as follows:

The target price of HK$260 is also based on 20 times the expected 2030 P/E ratio, discounted to December 2026 using a 15% weighted average cost of capital.