Tencent Holdings saw its Hong Kong-listed shares jump nearly 8% on Tuesday, with trading volume swelling to HK$17.7 billion, driven by two interconnected catalysts. Meta's AI agent application Muse has rapidly gained popularity in the United States, topping app store download charts, while Tencent unveiled its latest image generation model, Hy Image 3.5 Preview, on the same day. This has sparked a clear mapping narrative in the market—Tencent, with its super social ecosystem and high-frequency life scenario entry points, is seen as the most qualified candidate to serve as China's equivalent of Muse.

Meta's Muse climbed to the top three on the US App Store free charts within a day of launch and has consistently held a leading position in productivity rankings. Muse's core differentiator goes beyond the chatbot category: by reading social relationships and integrating APIs for email, calendar, shopping, food ordering, and more, it autonomously executes complex cross-application tasks for users in the background.

This viral success has redirected capital market attention toward domestic counterparts. WeChat, with its vast mini-program service ecosystem and closed-loop payment system, is considered to have efficiency and commercialization potential rivaling Meta's if AI agents were deeply embedded. Tencent thus became the most sought-after beneficiary stock that day.

Where the opportunity lies

A CICC research report noted that Meta Muse's breakthrough has shifted from "whether the model is smart enough" to "whether users can confidently delegate tasks and whether the product can connect to real-world transactions." The competitive focus for consumer-facing AI agents has officially moved from technology down to ecosystem distribution and trust architecture. Gary Tan, portfolio manager at Allspring Global Investments, said: "Some investors are drawing parallels between Tencent and Meta, particularly given WeChat's unique social ecosystem and its potential to support large-scale personal AI assistants."

Tencent accelerates with new image model

Tencent's release of Hy Image 3.5 Preview marks its latest move to compete head-on in the AI image generation space. The company stated the new model delivers improved performance over its predecessor and has already been integrated into Yuanbao as well as video editing and design tool product lines.

Tencent used several hundred internal designers as a testing group and claimed the model performs on par with ByteDance's Seedream 5.0 Pro, while slightly outperforming Alphabet's Google Nano Banana Pro and Alibaba's Qwen-Image-3.0 Pro. With thousands of game designers and artists on staff, Tencent is well-positioned as a primary beneficiary of the tool, though the company did not provide quantitative evidence for its quality claims.

The timing of the release, coinciding with Alibaba's AI conference opening on the same day, was notably conspicuous. Since hiring former OpenAI researcher Yao Shunyu as chief AI scientist, Tencent's AI strategy has undergone a shift: Yao has publicly criticized the practice of training models solely for benchmark scores, emphasizing product integration and solving real-world problems. Another former OpenAI researcher, computer vision expert Tian Yonglong, also joined Tencent's Hunyuan team in July to lead visual language model development.

The Muse playbook: from chat to delivery

Understanding the core of Tencent's rally requires grasping what new narrative Meta Muse has opened.

A CICC research report outlined four substantive differences between Muse and previous similar agent products. First, background asynchronous execution retains long-term memory and user preferences across conversations, allowing tasks to continue even after users close the app. Second, real account connections integrate email, calendar, payments, health, e-commerce, smart home, and other scenarios through built-in connectors, public APIs, and browser operations. Third, isolated execution with approval gating gives each user a dedicated Muse Secure VM runtime environment, with a Sentinel agent guarding all external operations, embedding security architecture as a core product design principle. Fourth, the Ideas and Feed features already hint at a recommendation-style agent prototype, with Muse shifting from waiting for user queries to proactively identifying needs and defining problems in advance.

The CICC report argued that the endgame for consumer-facing agents will resemble "recommendation" rather than "search"—agents proactively identify needs and solve problems for users based on sufficient context, rather than passively waiting for instructions. Meta's advantage lies in distribution and trust infrastructure: 3.6 billion daily active users provide scale, WhatsApp conversation threads lower the usage barrier, and long-term content behavior on Instagram and Facebook accumulates a richer personal needs profile that products like ChatGPT, which rely solely on chat history, find hard to replicate.

On the business model front, Muse adopts a free tier alongside two monthly subscription tiers at $20 and $100, with the highest tier approaching pricing levels of some B2B software. The CICC report suggested that as Muse handles shopping, booking, and other consumption tasks, its monetization path could evolve into a hybrid model of "subscription as a base plus transaction commissions."

Why Tencent is the optimal choice for a China version of Muse

The capital market's comparison of Tencent to Meta Muse is not a simple analogy but is supported by clear ecosystem logic.

When examining differences between US and Chinese consumer agent ecosystems, the CICC report noted that WeChat has integrated accounts, payments, and life services within its platform and possesses a unique mini-program ecosystem, allowing its own agents to directly call capabilities and complete transactions within authorized scopes. In contrast, Meta's overseas service entry points are relatively fragmented, with web pages still serving as an important carrier. Muse must rely on browser operations to cover long-tail services, resulting in higher connection and maintenance costs, and website redesigns could also impact execution efficiency.

In other words, if Tencent deeply embeds AI agents into the WeChat ecosystem, its natural mini-program closed loop and payment system would give it lower execution friction than Meta. Tencent's vision—building AI capable of executing various tasks for over one billion users in the WeChat ecosystem—aligns closely with Muse's product direction. This is the core logic behind the market's premium pricing of Tencent that day.

Timing determines pace: two paths and four stages

The CICC report introduced a product innovation cycle framework to divide the development of consumer-facing agents into four stages: the exploration stage (where both technical effectiveness and costs fail to meet thresholds), the transitional innovation stage (where technical effectiveness meets thresholds but costs remain high), the full innovation stage (where both effectiveness and costs break through thresholds), and the strong-get-stronger stage (where the technology curve flattens).

Using the historical paths of BlackBerry and iPhone as reference points, the report noted that the transitional innovation stage corresponds to BlackBerry's focus on enterprise email and trimming non-core functions, with BlackBerry's revenue growing from $600 million in 2004 to a peak of $20 billion in 2011. The full innovation stage corresponds to the iPhone's all-in-one integration, which ecologically crushed transitional products.

The CICC report believes that continuous AI model capability improvements and declining token computing costs are relatively certain factors going forward. What is harder to predict are users' tolerance for agent task errors, their willingness to shift existing habits, and their cost thresholds. Based on this, consumer agents face two possible development scenarios: one where they first undergo a transitional innovation phase, starting from general office needs and gradually extending to comprehensive scenarios; another where technology and cost curves decline rapidly in tandem, allowing a direct leap past the transitional phase into full innovation—in which case, giants with distribution advantages and ecosystem depth like Meta and Tencent would benefit first. The CICC report emphasized that the shorter the interval from transitional to full innovation phases, the greater the advantage for internet giants; conversely, startups would have more time to build moats.