Spring Festival First Battle, Alibaba AI "Aircraft Carrier" Begins to Take Shape

On New Year’s Eve 2026, Alibaba officially open-sourced the next-generation large model Qwen3.5-Plus, praised by AI enthusiasts as the “Most Powerful Source God.” Compared to previous generations of Qwen large language models, Qwen3.5 has achieved a generational leap from a pure text model to a native multimodal model.

In the early hours of the first day of the Lunar New Year, data released by Alibaba’s Qwen App during the Spring Festival showed that over 130 million people nationwide experienced AI shopping for the first time, with a total of 5 billion “Qwen Help Me” requests, making Qwen an everyday AI assistant for the public.

From open-sourcing top-tier large models to full-ecosystem application deployment, Alibaba AI is experiencing dual-line explosive growth. This Spring Festival, Alibaba proved with action that AI can reach for the stars and the sea, as well as protect everyday life.

Qwen: Achieving National-Level AI Assistant Status

How to enhance the commercialization of AI applications for end users has become a new challenge faced by global tech companies.

If last year’s Spring Festival was when DeepSeek gained recognition through deep thinking, this year, Qwen made a strong breakthrough with “AI handling tasks”—“Qwen Help Me” is becoming a new user habit in the AI era.

On February 6, Qwen launched its Spring Festival campaign. During this period, over 130 million people nationwide used AI for shopping for the first time, with users making 5 billion “Qwen Help Me” requests. Qwen officially entered the ranks of national-level AI assistants. It marked the first large-scale real-world scenario execution and commercialization validation of AI Agents globally, achieving a deep integration of “instructions” and “transactions,” reshaping how users interact with the internet. This not only makes AI truly accessible to everyone but also represents a significant milestone in AI development history.

Latest data from QuestMobile shows that Qwen’s daily active users (DAU) reached 73.52 million, far surpassing Yuanbao’s 18.28 million and approaching Doubao’s 78.71 million. This means that in less than two months, Qwen has matched the user base accumulated by industry-leading products over three years.

Alibaba data indicates that in the first two days of the Spring Festival, user orders for AI ticket purchases increased 22-fold month-over-month, and AI bookings for flights and transportation tickets grew over 7 times; during the Spring Festival, AI movie ticket orders surged 372 times month-over-month, with AI orders in third- and fourth-tier cities skyrocketing by 782 times.

Notably, nearly half of AI orders came from county markets. The convenience of completing operations with a single command allowed nearly 4 million users over 60 years old to try AI shopping for the first time.

Leveraging Alibaba’s ecosystem, Qwen can seamlessly coordinate services across platforms like Taobao, Alipay, Taobao Flash Sale, Fliggy, Amap, and Damai. Future plans include launching features such as AI ride-hailing, mobile phone recharge, and Amap street sweep group buying.

Targeting General Artificial Intelligence

On New Year’s Eve, Alibaba open-sourced the new generation large model Qwen3.5-Plus, which matches the performance of Gemini-3 Pro and tops the global list of the most powerful open-source large models. Qwen3.5 has undergone a comprehensive overhaul of its underlying architecture. The released Plus version has a total of 397 billion parameters, with only 17 billion active parameters, achieving “big performance” with “small parameters,” outperforming Qwen3-Max with over a trillion parameters.

Alibaba open-sourced Qwen3.5-Plus, with performance comparable to Gemini-3 Pro.

From Qwen2.5 and Qwen3 to Qwen3.5, the past two years have seen multiple breakthroughs in the capabilities of open-source large models, all achieved by Qwen. Currently, Qwen has surpassed 1 billion downloads worldwide, with monthly downloads exceeding the combined total of ranks 2 through 8, and over 200,000 derivative models created by developers, making it the world’s leading open-source large model.

Industry experts say that during this Spring Festival, Qwen revealed its true killer feature.

Unlike previous pure text models, Qwen3.5 has achieved a generational leap from text to native multimodal capabilities. The model significantly expanded training data in Chinese, English, multiple languages, STEM, and reasoning, leading in benchmarks across reasoning, programming, and agent intelligence, even surpassing closed-source top products like GPT-5.2 and Gemini-3 Pro.

Qwen3.5 is just the beginning of Alibaba’s multimodal layout. Future focus will be on all-in-one multimodal integration, striving toward AGI (Artificial General Intelligence). Currently, Alibaba is among the world leaders in multimodal understanding and generation.

A week ago, Qwen released Qwen-image-2.0, integrating image generation and editing, ranked among the top three globally on the AI Arena leaderboard, dubbed the “Nano Banana Pro” of domestic AI. Now, with further breakthroughs in visual understanding, Qwen3.5 advances toward full multimodal capabilities.

Industry observers believe that the development of models and multimodal routes are not isolated but are building toward full multimodal technology. Alibaba aims to evolve large models from single-task processing to full-scenario understanding and generation of text, vision, and speech, truly achieving “comprehend, listen clearly, express, and execute,” ultimately moving toward artificial general intelligence.

A Giant Tech Company in the Making

Qwen3.5-Plus API pricing starts as low as 0.8 yuan per million tokens, only 1/18 of Gemini-3 Pro at comparable performance, backed by Alibaba’s deep collaboration and innovation in AI computing power and models.

At the same time, Alibaba boasts the most comprehensive and rich ecosystem and fulfillment capabilities in China, covering all scenarios from dining, entertainment, travel, and accommodation. AI has further integrated into Alibaba’s ecosystem, bringing ultimate convenience in the AI era.

Behind this dual-line AI surge is Alibaba’s accumulated strength. The “Tongyun Ge”—a golden triangle formed by the DAMO Academy, Alibaba Cloud, and Pingtouge—gives Alibaba control over core AI computing power.

On January 22, multiple media reported that Alibaba plans to support Pingtouge, a chip company, for future independent listing. Industry insiders see Pingtouge as a key strategic move in Alibaba’s AI layout, with chips, cloud, and models forming the “black land” for Alibaba’s long-term AI development.

As the “Tongyun Ge” golden triangle takes shape, Alibaba has become one of the few global tech giants, and the only Chinese company, to achieve full-stack vertical integration of “self-developed AI chips + world-class cloud platform + world-class large models.”

In the global AI race, only vertical integration can achieve ultimate optimization, which is the core logic behind Alibaba’s creation of “Tongyun Ge.”

This layout was not built overnight: as early as the 2017 Cloud Computing Conference, Alibaba announced the establishment of DAMO Academy to focus on chips and cutting-edge technology; in 2018, Pingtouge was founded, simultaneously launching large model research. Before the AI explosion, Alibaba had already made saturated investments in foundational technologies.

Google demonstrated the value of full-stack integration with the “Gemini model + TPU chips + cloud platform,” but Alibaba is surpassing this model. With self-developed chips, Alibaba can customize “Zhenwu” chips for the Qwen architecture and optimize resource scheduling through “the most chip-savvy cloud.” Alibaba Cloud is no longer just a traditional cloud platform but a “AI supercomputer” combining hardware and software.

With world-leading cloud infrastructure, independently developed high-end AI chips, and the strongest open-source large models globally, Alibaba in the future will become a giant tech company building the “water, electricity, coal” infrastructure for the AI era.