<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/74fd7011-a6bb-4f4c-a748-233f8fe7403f.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="Image" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/74fd7011-a6bb-4f4c-a748-233f8fe7403f.png"/>
The &#34;Happy Family&#34; of Alibaba ATH has gained a new member.
Just after HappyHorse surged onto global rankings with its video generation capabilities, pulling Alibaba back to the center stage of multimodal AI, Alibaba quickly played another card.
On April 16, 2025, Alibaba released the &#34;Open World Model Product HappyOyster.&#34; While the name continues the HappyHorse system, the direction is clearly different.
If HappyHorse solves &#34;how to generate better videos,&#34; HappyOyster attempts to answer &#34;how to generate a world that can persist continuously and interact in real-time.&#34;
Coincidentally, on the same day, Tencent also officially released and open-sourced its Hunyuan 3D &#34;world model&#34; HY-World2.0, focusing on 3D scene generation importable into game engines and the construction of interactive worlds.
Almost simultaneously, two Chinese internet giants bet on the narrative of &#34;world models,&#34; which in a sense means AI competition has entered the next phase: moving from answering questions and generating text to building spaces, simulating reality, and creating interactive digital worlds.
<h2>Playing the Card</h2>
Over the past year, the AI video generation industry has formed a familiar public process: input a prompt, wait tens of seconds or minutes, and receive a finished video clip lasting from a few seconds to dozens of seconds. Users act as directors and clients alike, providing requirements and waiting for the machine to deliver results. This model addresses content production efficiency but remains essentially one-way generation.
HappyOyster attempts to break this logic.
Alibaba defines it as a &#34;world model&#34; because it allows users to issue continuous instructions during the generation process, with visuals responding in real-time and the world evolving continuously. You can move your viewpoint within an already generated space, change the environment, schedule characters, and rewrite the plot. The model does not deliver results once; it runs continuously. Users are no longer just prompt inputters but become participants, controllers, and even co-creators of this world.
However, strictly speaking, HappyOyster is still built upon a generative model framework and is not the complete, self-consistent, general-purpose world model with high physical laws and causal logic that outsiders imagine.
It is more like a transitional form: taking a step forward from video generation toward real-time simulation and continuous interaction. Visual fidelity remains limited, there is latency in movement, and character consistency and instruction understanding need improvement, all indicating it is still in the early stages.
But what matters is never whether it is perfect today, but that it has clarified the direction.
In the past, when discussing AI, most attention focused on large language models. Whose parameters were larger, whose reasoning was stronger, who wrote code more accurately, and who could replace office software. After entering 2026, this narrative began to show diminishing marginal returns.
Text capabilities continue to improve, but user novelty is fading, and capital markets increasingly understand: simple chatbots are hard to support high valuations. What truly opens new markets is multimodality, video, interaction, and digital worlds.
Alibaba clearly perceived this.
The sequential launch of HappyHorse and HappyOyster appears to be two products, but they are actually two moves on the same strategic path. The first step proves its global competitiveness in video generation and regains market attention; the second elevates the competitive dimension from video generation to world generation, upgrading Alibaba's AI narrative from &#34;can build models&#34; to &#34;can define the next generation of interaction forms.&#34;
Behind this lies a microcosm of Alibaba's significantly accelerated pace in the AI field recently.
Not long ago, Alibaba completed an internal AI architecture restructuring, establishing the ATH (Alibaba Token Hub) business group, integrating the Tongyi Lab, MaaS business line, Qianwen division, Wukong division, and AI Innovation division into a unified system.
At the time, outsiders focused more on organizational moves, but the true significance lies in Alibaba regrouping resources previously scattered across cloud, models, applications, and innovation teams with a single goal: putting AI product development and commercialization on a fast track.
HappyOyster was born against this backdrop.
It comes from the Innovation Division under ATH, belonging to the same team as HappyHorse. This point is quite interesting. In recent years, Alibaba's most well-known assets in AI were Tongyi Qianwen and Alibaba Cloud; innovative businesses were often seen as marginal test fields.
Now, however, Alibaba's most radical and talk-worthy products come precisely from the Innovation Division. This indicates a new mechanism forming within Alibaba: mature businesses handle steady growth, while innovation teams bet on future entry points.
This division of labor resembles Alibaba's past approach in e-commerce. After Taobao matured, Tmall took over branding, Xianyu handled the long tail, and Hema explored offline retail.
Today, Alibaba has brought this logic to the AI era: Tongyi handles foundational models, Alibaba Cloud handles commercialization, Qianwen handles mass-market entry points, and products like HappyHorse and HappyOyster undertake the task of exploring future interaction paradigms.
<h2>Major Shock</h2>
From an industry perspective, the emergence of HappyOyster carries another layer of meaning: it marks the beginning of direct confrontation among domestic tech giants in the interactive generation sector.
Tencent's newly launched HY-World2.0 emphasizes industrial implementation more strongly. It can generate assets in multiple formats such as Mesh, point clouds, and 3DGS, and connects with existing game engine workflows, facilitating developers to quickly build maps, levels, and prototype scenes.
Tencent's route is clear: let the model first serve its strongest gaming business, solving problems of long development cycles for open-world games, high art costs, and insufficient content capacity.
Behind this lies Tencent's very realistic calculation.
As the company with the highest global gaming revenue, Tencent possesses massive 3D data, mature engine experience, and continuous content demands. If the model matures, the R&amp;D system for games will benefit first. Maps that previously took months to draft might be generated with a single sentence in the future; prototyping that previously required dozens of artists might be completed by a few people plus AI.
Alibaba's path differs slightly. HappyOyster emphasizes real-time interaction, world evolution, and open creative experiences. Unlike Tencent, which cuts into the industry toolchain first, it is closer to the next-generation user interaction entry point. Tencent aims to solve production problems first, while Alibaba tests the future form of consumer-grade experiences.
Neither route is superior; they simply start from different points. Tencent enters from the game industry, while Alibaba enters from content interaction, but the destination may be the same place: AI becoming the real-time generation engine for the digital world.
This also indicates that the logic of Chinese tech giants in AI competition has changed. The previous phase compared whose general model parameters were larger and whose benchmark scores were higher; this phase compares who can find the deepest landing directions aligned with their own businesses. Tencent bets on 3D worlds because gaming is a cash cow; Alibaba bets on multimodal interaction because e-commerce, content, and cloud computing all require richer digital space entry points.
From this perspective, even if HappyOyster is not yet mature, it deserves serious consideration. What it reveals is not product completion but corporate directional sense. A giant willing to invest resources in such highly uncertain frontier products itself indicates that internally it has judged: the next phase of AI is not just generating content but generating environments; not just answering questions but carrying behaviors.
For Alibaba, this step is particularly crucial.
In recent years, discussions about Alibaba often focused on e-commerce competition, slowing cloud growth, and frequent organizational adjustments, while the AI narrative was overshadowed by ByteDance, Tencent, Baidu, and even startups.
But in recent months, Alibaba has clearly attempted to regain the initiative. Wu Yongming explicitly set a goal for cloud and AI annual revenue to exceed $100 billion in the next five years, elevated AI to a core position within the organization, and densely released new models and forms in products.
HappyHorse was a shot proving technical strength; HappyOyster is a shot proving strategic ambition.
Because video generation remains competition in a stock market, while interactive generation represents an incremental market. It connects not only content creation but potentially games, education, cultural tourism, virtual social networking, robot training, and even future spatial computing device interaction systems. Whoever takes a leading position at this stage has the opportunity to participate in defining the next generation of computing platforms.
Alibaba clearly does not want to be merely a model supplier within someone else's ecosystem.
Today's HappyOyster may still be a somewhat rough &#34;Happy Oyster,&#34; and Tencent's HY-World2.0 is still in early polishing stages, but the fact that two giants played cards on the same day illustrates a reality: interactive generation is no longer just a laboratory concept but has begun to become a battleground for Chinese technology companies.
The real competition has just begun.

00700

TCEHY

TCTZF

BABA

CLOU

DTCR

KBAB

BABO

BABX

The Battle for the Next-Generation Entry Point Begins with 'Building Worlds'

- Alibaba's new product, HappyOyster, was launched on April 16, 2025, aiming to create a continuously evolving and interactive digital world, distinct from its predecessor, HappyHorse, which focused on video generation.  
- Meanwhile, Tencent introduced its own 3D world model, HY-World2.0, underscoring a shift in the AI competition towards building interactive environments rather than merely generating content.  
- This development signals a pivotal moment for Chinese tech giants as they strive to redefine user interaction and capitalize on emerging market opportunities.

Wallstreetcn

TENCENT

Tencent

Tencent Holdings Limited

Alibaba

Global X Cloud Computing ETF

Global X Data Center & Dgtl Infrs ETF

KraneShares 2x Long BABA Daily ETF

YieldMax BABA Option Income Strategy ETF

BABA 2x Long Daily ETF - GraniteShares

Alibaba and Tencent Play Their Cards on the Same Day as AI Begins to Compete for the 'World Entry Point'