As 2025 dawned, Alibaba Cloud was abuzz with activity, first announcing a strategic partnership with Zero One Everything in model platform business and then integrating LeTV's innovative AR glasses with Tongyi Large Model. As the cloud vendor with the highest market share in domestic cloud computing, Alibaba Cloud's development has remained steady. Meanwhile, the Tongyi series of products, as Alibaba Cloud's primary native large model, has also progressed normally in To B collaborations over the past year.

However, towards the end of 2024, Alibaba Cloud spun off the Tongyi Large Model's To C team, including product managers in the Tongyi To C direction and related engineering teams, which were subsequently integrated into Alibaba's Intelligent Information Business Group, alongside Quark. This can be seen as Alibaba setting a new direction for the Tongyi To C team's development in 2025.

2024 was a year when major vendors' large model To C applications were all striving to make their mark. According to Qimai's 2024 AI Product Download List Top 30, the top three chatbot apps were Doubao, Kimi, and Wenxiaoyan, with Tongyi ranking fourth, with a significant gap in downloads compared to Doubao.

In theory, Tongyi is not lacking in powerful large models, especially in the open-source domain. However, its slow progress in To C in 2024 was evident to all.

This does not mean that Alibaba neglected AI To C in 2024; the Tongyi App has been consistently updated over the past year. Nor does it imply that Alibaba is not adept at AI To C; after all, Quark within the Alibaba ecosystem has achieved remarkable success. In Qimai's 2024 annual download list, Quark surpassed Doubao to rank first.

Therefore, from the perspective of "New Standpoint," if we were to summarize the state of Tongyi To C in 2024, "marginalization" might be a more fitting term than "indecision".

Tongyi To C's marginalization was not due to Alibaba Cloud or Alibaba's AI strategy on the surface, but rather to Alibaba's own organizational structure.

The specific target of marginalization is the "Tongyi" IP (brand effect category). As a C-end product, it has failed to cultivate its influence on other businesses within Alibaba's AI ecosystem over the past year, as no one within Alibaba needed it. Consequently, its influence on C-end users has also not been well cultivated.

01. How the Best Logical Chatbot of 2024 Was Marginalized

The model capabilities that the Tongyi app can invoke are naturally not inferior to those of other domestic chatbot apps.

For example, the Qwen2.5-72B-Instruct released last year was once the only Chinese large model in the top ten of the Chatbot Arena list for a period after its release.

There are numerous AI benchmarks, and over the past one to two years, it is not uncommon for large model vendors to use benchmark lists favorable to themselves to support their claims. However, due to its mechanism, Chatbot Arena currently appears to have a certain degree of authority: it allows anyone to ask questions to two random anonymous models and vote for the answer they prefer. Compared to some other benchmarks, the Arena test with its randomness and anonymity seems more objective.

Although at the beginning of 2025, the top Chinese large model on the Chatbot Arena list has been replaced by DeepSeek-V3, at least in 2024, the Tongyi Qwen2.5 series still held a leading position in capability evaluations.

Of course, the non-profit organization behind Chatbot Arena is related to many large model vendors, and the users participating in the evaluations are most likely those who care about this list. The objectivity of Chatbot Arena may not be absolute.

But regardless of which large model of the Tongyi series the Tongyi app currently invokes, when we tested it with a previously non-existent philosophical dialectical question, the author subjectively perceived that the answer from the Tongyi app was more complete and dialectical in its argumentation compared to Doubao, Kimi, and Wenxiaoyan.

Moreover, the DeepSeek app's answer to this question was also superior to that of the Tongyi app, which is consistent with the results of the Chatbot Arena list. (The specific testing process and comparison of results will be provided in the next article on the Tongyi product level)

When multiple and sufficient proportions of subjective perceptions point to the same result, that result becomes "objective". Therefore, we can still conclude that in the time dimension of 2024, the Tongyi app was probably the best logical chatbot in China.

Yet, despite being such a powerful large model application, its development in To C lagged behind this year.

According to Qimai's 2024 AI Product Download List Top 30, the download rankings of chatbot-type AI products were Doubao (2nd), Kimi (7th), and Wenxiaoyan (9th).

Among them, Doubao's cumulative downloads reached over 200 million, while Tongyi missed the top three in chatbot apps, ranking fourth (18th) with downloads in the range of over 20 million.

However, it is not that Alibaba neglected AI To C in 2024.

The Tongyi App has been updated consistently over the past year. Even from the perspective of the iOS App Store, its version update frequency is comparable to that of Doubao. As of now (early 2025), the update cycle length for Tongyi and Doubao apps over the past 25 versions is 6 to 7 months; Wenxiaoyan has the lowest update frequency at 9 months; and Kimi has the highest at 3 months.

Nor is it that Alibaba is not adept at AI To C now, considering that another product within the Alibaba ecosystem, Quark, achieved remarkable success in 2024. In Qimai's 2024 annual AI app download list, Quark surpassed Doubao to rank first.

This can perhaps be summarized as Alibaba's uncertain path in Tongyi To C in 2024, but this conclusion holds only if Alibaba's layout strategy in the AI era is solely responsible for Tongyi To C.

However, this is not the case. Alibaba's system determines that it will not have a highly consistent layout for its internal AI ecosystem, with each plate basically operating independently and being responsible for its own profits and losses. Therefore, from the perspective of "New Standpoint," if we were to summarize the state of Tongyi To C in 2024, "marginalization" might be a more fitting term than "indecision".

As mentioned at the beginning, as 2025 dawned, Alibaba Cloud first announced a strategic partnership with Zero One Everything in model platform business, and then AR glasses were integrated with Tongyi Large Model.

There are two pieces of information that can be basically confirmed here: first, Alibaba Cloud will not only support Tongyi as its large model; second, even if the Tongyi series is the large model product with the most resources倾斜 from Alibaba Cloud, due to the overall Alibaba structure, Tongyi's current business achievements at the To B level have little to do with the Tongyi To C team.

This means that the Tongyi app has not been able to synergize more ecological resources within its parent company like Doubao, but Tongyi itself could have.

02. From Watching Others "Fight Independently" to "Fighting for Oneself"

To understand this situation, a comprehensive analysis of the positions of various parties within Alibaba is needed.

From the perspective of Alibaba Cloud, where the Tongyi To C team was previously located, there was no demand or necessity to create a good AI To C product.

It is evident from LeTV's AR glasses accessing Tongyi Large Model through Alibaba Cloud that after the Tongyi To C team was spun off from Alibaba Cloud, major B-end partners can still connect with the superior Tongyi Large Model through Alibaba Cloud.

For these partners, "Alibaba Cloud" is the influential brand, while "Tongyi" is just one of its product series.

On the other hand, as a cloud vendor, Alibaba Cloud sells cloud services. It is one of the mainstream narratives in the current market for cloud vendors to cooperate with more large model vendors, such as the strategic partnership with Zero One Everything in model platform business, and the cloud services behind other unicorn large models are also provided by Alibaba Cloud.

When a large company has already developed a robust cloud business with a high market share before the explosion of large models, the influence of its flagship large model and To C products will relatively be overshadowed.

The same situation also occurred with Tencent Cloud, while in contrast to Alibaba Cloud and Tencent Cloud, ByteDance's Volcano Engine stands out. The "influencer" role of its flagship product, Doubao, within Volcano Engine and other products within the ByteDance ecosystem, may be the "outlier" among cloud vendors.

Additionally, from Alibaba's overall perspective, in the AI era, Alibaba already has a more powerful product, Quark, to compete with Doubao. From the perspective of other businesses within Alibaba, under the influence of its organizational structure, their main goal is to survive and profit on their own.

Alibaba Cloud does not need the Tongyi app to drive sales, nor do other products within Alibaba. For example, supporting the Quark AI system is Quark's self-developed large model, and Tmall Genie is integrated with Quark.

Generally speaking, the mutual selection criteria for cooperation between an AI business and ecological partners are: software products need cloud services for AI, hardware products need large models for AI; strong products need the technology itself for AI, while weaker products need the influence of AI.

This selection benchmark applies to any enterprise. After combining this benchmark based on one's own product positioning, various selection results can be traced. When this benchmark is applied to Alibaba's structure, it results in the current situation.

Only after understanding the fundamental reason for Alibaba's structure marginalizing Tongyi To C can we better comprehend the adjustment of the Tongyi To C team at the end of the year.

Now the Tongyi To C team has been incorporated into Alibaba's Intelligent Information Business Group, which also includes content-based products such as Quark, UC, and Shuqi. The core model of these products is "intelligent tools + content + services". This model is also suitable as a framework for the independent development of the Tongyi To C application.

Previously, the essential reason for Tongyi To C being marginalized was that the AI-related sectors within Alibaba were "fighting independently". The current solution is naturally to let Tongyi To C detach from Alibaba Cloud and "fight for itself".

"The general trend of the world is that after a long period of unity, there must be separation, and after a long period of separation, there must be unity." At the end of last year, Alibaba established the Alibaba E-commerce Business Group, comprehensively integrating e-commerce businesses such as Taobao and Tmall Group, Alibaba International Digital Commercial Group, as well as 1688 and Xianyu, forming a business cluster covering the entire domestic and foreign industrial chain, with Jiang Fan serving as CEO.

Perhaps it will be difficult to integrate all AI-related businesses within Alibaba like e-commerce businesses. However, how to maintain the external competitiveness of each business while ensuring internal collaborative development? Exploring the AI business structure will be a main topic for Alibaba's overall future for a long time to come.

03. Final Thoughts

2024 is now in the past, and for the Tongyi To C team, the key concern is whether the "delay in the battle" over the past year has had a significant impact.

"New Standpoint" believes the impact is limited. Take DeepSeek as an example; there are still new chatbot apps emerging with the fame of its eponymous large model, and the vast majority of users have not yet formed a consistent perception of "who is the best chatbot". The large model market is still in a stage of high-intensity shuffling.

Moreover, in 2024, both Doubao and Kimi were still in a stage where user acquisition through advertising was far greater than that through the capabilities of the products themselves.

At the end of the year, information from a research report by Zhejiang Merchants Securities was disclosed. The report pointed out that ByteDance's capital expenditure on AI in 2024 reached 80 billion yuan, close to the combined total of Baidu, Alibaba, and Tencent (about 100 billion yuan).

What money can achieve is not a moat for large companies vis-à-vis each other.

And if we were to find a benchmark application to guide the development of the Tongyi app in the coming year, Quark from the same sector would also be a good fit. Regarding this point, "New Standpoint" will provide specific reasons in the following article, "If Not Tech-Centric, How Can Tongyi To C Refer to Quark and Doubao?"

In 2025, the Tongyi app still has many opportunities.

*The title image and images in the text are sourced from the internet.