Alibaba and Tencent make moves on the same day; AI begins competing for the "world gateway."

```

The "Happy Family" of Alibaba ATH has gained a new member.

Just after HappyHorse made it onto global charts with its video generation capabilities and brought Alibaba back to the center stage of multimodal technology, Alibaba quickly played another card.

On April 16, 2025, Alibaba launched the “open world model product HappyOyster.” The name seems to follow the HappyHorse system, but the direction is clearly different.

If HappyHorse addresses the question of “how to generate better videos,” HappyOyster tries to answer “how to generate a world that can persist and can interact in real-time.”

Coincidentally, on the same day, Tencent officially released and open-sourced its Hunyuan 3D “world model” HY-World2.0, focusing on generating 3D scenes for game engines and building interactive worlds.

Both Chinese internet giants placing bets on “world models” at almost the same time, to some extent, signals that AI competition has entered the next stage: from answering questions and generating text, to building spaces, simulating reality, and creating interactive digital worlds.

Playing the Card

Over the past year, the AI video generation industry has formed a process familiar to the public: input a prompt, wait a few seconds or minutes, and get a video of a few seconds to tens of seconds. Users act like directors and clients, giving requirements and waiting for machines to deliver results. This solves content production efficiency but is still essentially one-way generation.

HappyOyster is attempting to break this logic.

Alibaba defines it as a “world model” because it allows users to continuously give commands during the generation process, with real-time visual feedback and continual world evolution. You can move the viewpoint, change the environment, manage characters, and rewrite the plot within an already generated space. The model isn’t a one-time deliverable—it keeps running. Users are no longer just prompt inputters, but participants, controllers, and even co-creators of the world.

But strictly speaking, HappyOyster is still built on generative model architecture. It isn’t yet the kind of complete, self-consistent, general world model with high levels of physical rules and causal logic imagined by the outside world.

It’s more like a transitional form: advancing from video generation toward real-time simulation and continuous interaction. Visual detail is still limited, there are delays in movement, and character consistency and command comprehension need improvement, suggesting it is still in its early stages.

But what matters is not whether it’s perfect today, but that it has a clear direction.

In previous AI industry discussions, most attention has been on large language models: who has more parameters, stronger reasoning, writes better code, or can replace office software. After entering 2026, this narrative is reaching diminishing returns.

Text abilities are still improving, but user excitement is waning, and the capital market is increasingly aware: pure chatbots struggle to sustain high valuations. What truly opens new markets are multimodality, video, interactivity, and digital worlds.

Alibaba is clearly aware of this.

The launches of HappyHorse and HappyOyster are not independent—they are sequential moves on the same strategic path. Step one is to prove global competitiveness in video generation and regain market attention; step two is to raise the competitive stakes from video generation to world generation, upgrading Alibaba’s AI narrative from “can make models” to “can define the next generation of interactive formats.”

Behind this is a microcosm of Alibaba’s obvious recent acceleration in AI.

Not long ago, Alibaba completed a major internal AI reorganization, establishing the ATH (AlibabaTokenHub) business group, unifying Tongyi Lab, the MaaS business line, Qianwen Division, Wukong Division, and AI Innovation Division.

At the time, the outside world focused more on the organizational changes, but the real significance was bringing together previously scattered resources in cloud, models, applications, and innovation teams, with one goal: to put AI product R&D and commercialization into the fast lane.

It’s in this context that HappyOyster was born.

It comes from ATH’s Innovation Division, the same team as HappyHorse. This is noteworthy. In recent years, Alibaba’s best-known AI assets have been Tongyi Qianwen and Alibaba Cloud, with innovation often seen as a marginal experimental field.

But now, Alibaba’s most aggressive, headline-grabbing products come from its Innovation Division. This indicates that internally a new mechanism is forming: mature businesses drive steady growth, while innovation teams bet on future opportunities.

This kind of division of labor is much like how Alibaba did e-commerce in the past. After Taobao matured, Tmall took over branding, Xianyu handled the long tail, and Hema explored offline retail.

Now Alibaba is applying this logic to the AI era: Tongyi handles foundational models, Alibaba Cloud oversees commercialization, Qianwen manages mainstream user access, while products like HappyHorse and HappyOyster explore the next generation of interactive paradigms.

Earthquake

From an industry perspective, the appearance of HappyOyster has another meaning: it pushes domestic major companies to truly compete head-on in the interactive generation track.

Tencent’s HY-World2.0 emphasizes industrial application. It can generate a variety of asset formats like Mesh, Point Cloud, and 3DGS, and links with existing game engine workflows, making it easy for developers to quickly build maps, levels, and prototype scenes.

Tencent’s path is clear: use models firstly in its strongest gaming business, addressing issues of long open-world development cycles, high art costs, and insufficient content production.

There is a very realistic calculation behind this.

As the company with the world’s highest gaming revenue, Tencent possesses vast 3D data, mature engine experience, and a continuous demand for content. Once the models mature, the biggest beneficiary is the game development system. Maps that used to take months to prototype may one day be generated with a prompt; prototype art that used to require dozens of artists might soon be completed by a few people with AI.

Alibaba’s approach is somewhat different. HappyOyster focuses more on real-time interaction, world evolution, and an open creative experience, closer to the next-gen user interaction entry point than the industry toolchain focus of Tencent. Tencent wants to solve the production problem first, while Alibaba is exploring the future shape of consumer-grade experiences.

There is no superior route here, just different starting points. Tencent’s point of entry is game industry, Alibaba’s is content interaction, but the end point may be the same: AI as the real-time generative engine for digital worlds.

This also shows that the logic of Chinese big tech’s AI competition has changed. The previous stage was about who had the model with more parameters or better benchmarks; now it’s about who can find the landing direction most synergistic with its core business. Tencent bets on 3D worlds because games are its cash cow; Alibaba bets on multimodal interaction, because e-commerce, content, and cloud all need richer digital space entry points.

From this angle, even if HappyOyster is still immature, it should be taken seriously. What it reveals is less about product completeness and more about company direction. A big player willing to commit resources to such high-uncertainty, cutting-edge products is already an internal judgment that the next stage of AI is not just content generation, but environment generation; not just answering questions, but enabling behavior.

For Alibaba, this step is especially crucial.

In recent years, outside discussion of Alibaba has often focused on e-commerce competition, slowing cloud growth, frequent organizational changes, with AI narrative taken by ByteDance, Tencent, Baidu, or even startups.

But in the past few months, Alibaba is clearly working to regain the initiative. Daniel Zhang has set a clear goal of cloud and AI exceeding $100 billion in annual revenue in five years, organizationally raising AI to a core group position, and releasing new models and product forms at a rapid pace.

HappyHorse is the shot proving technical strength, HappyOyster the shot proving strategic ambition.

Because video generation is still about competing in an existing market, while interactive generation represents an incremental market. It connects more than just content creation—potentially games, education, culture and tourism, virtual social, robotics training, even the OS of future spatial computing devices. Whoever secures a position at this stage has the opportunity to help define the next computing platform.

Alibaba clearly doesn’t want to just be a model supplier in other people's ecosystems.

Today’s HappyOyster might still be a somewhat rough “Happy Oyster,” and Tencent’s HY-World2.0 is also still being polished, but both giants playing their cards on the same day already proves one fact: interactive generation is no longer just a lab concept, but is becoming a battleground for Chinese tech companies.

Real competition is just beginning.

Risk Disclosure and DisclaimerThe market carries risks, and investment should be approached with caution. This article does not constitute personal investment advice, nor does it take into account the particular investment objectives, financial situations, or needs of any individual users. Users should consider whether any opinions, perspectives, or conclusions in this article are appropriate for their individual circumstances. Investments based on this are at your own risk. ```