Doubao is going to charge fees too.
On May 4th, Doubao quietly updated its statement regarding paid version services on its App Store page. The statement said that in order to better serve professional users, Doubao will launch a paid subscription system that includes more value-added services while retaining the free version, and disclosed three pricing tiers: Standard Edition ¥68/month, Enhanced Edition ¥200/month, and Professional Edition ¥500/month.
Doubao has always provided free services. On the basis of free services, Doubao is also exploring the launch of more value-added services to meet the differentiated needs of different users. The details of the relevant plans are still in the testing phase, and complete information will be released through official channels when it is officially launched.
In other words, the functions that users can currently use Doubao for free, such as information research, basic copywriting, daily Q&A, and learning tutoring, will still be available for free in the future. The core logic of the paid version is to provide "value-added services" and will not affect the daily user experience of ordinary users.
This strategy is not new. Leading global AI products such as ChatGPT and Claude have all adopted a tiered model of "basic free + advanced paid": first, use free functions to penetrate users' minds and cultivate their habits, and then use advanced capabilities to serve core users who are willing to pay for value.
As for the price of 500 yuan/month, is it expensive or not? Looking at the numbers alone, the 500 yuan/month price for the professional version does indeed set a new price ceiling for general-purpose AI assistants in China. However, placing it within the global AI product pricing system yields a completely different conclusion. Combining the paid pricing of mainstream global AI products in May 2026, we can draw the following comparisons: First, the 68 yuan standard version is only less than 10 yuan higher than Wenxin Yiyan and Xunfei Xinghuo, basically hitting the mainstream benchmark for paid AI in China and not falling outside the acceptable range for the general public. Second, the 200 yuan enhanced version is in the same price range as ChatGPT Plus and Claude Pro's 145 yuan/month fee, targeting the mainstream paid tiers of leading global AI products and catering to users with high-frequency productivity needs. Third, the 500 yuan professional version is indeed the first time a domestically produced general-purpose AI product has reached the high-end price range, but it hasn't touched the global pricing ceiling of ChatGPT Pro and Claude Max. Essentially, it's a test of the willingness of heavy professional users in China to pay. However, the price is only related to the degree of value matching. 500 yuan/month might just be the cost of a month's coffee for many people, or a typical meal for two. For content creators, programmers, and small and medium-sized business owners who frequently interact with AI and rely on it for core production work, the cost of a 24/7 online, all-around assistant capable of writing, calculating, and analyzing data is far lower than hiring a human. But for ordinary users who only occasionally look up information or write emails, even the 68 yuan standard version is not worth the money.
Entering at this point, ByteDance has calculated two key issues
Why now?
Before discussing whether Doubao's paid subscription model can succeed, the more crucial question is—why did ByteDance choose this particular time to enter the market?
After all, ByteDance hasn't been quick to act in the paid subscription arena. Players like Baidu Wenxin Yiyan, Moon's Dark Side Kimi, and Xunfei Xinghuo launched mature subscription systems as early as 2024, while Doubao didn't officially launch until 2026. Perhaps it's because at this moment, ByteDance has calculated two key issues.
The first issue is the growth issue: the marginal benefit of trading scale for free services has almost reached its limit.
The first issue is the growth issue: the marginal benefit of trading scale for free services has almost reached its limit.
According to QuestMobile data, as of the end of March 2026, Doubao's monthly active users had exceeded 345 million, firmly ranking first among domestic C-end AI applications, covering a full range of users from students and professionals to creators and small and medium-sized enterprise owners. The domestic internet user base is only so large; the users who could potentially engage with AI have basically already been reached. Relying on a completely free model to spend money is no longer effective in attracting new users; instead, it incurs real computing costs for each new conversation. The second factor is the market: the user education for paid AI in China has been completed. When large-scale AI models first emerged in 2023, domestic users' understanding of AI was still that of a "novel toy," and they had a strong resistance to paying for it; but by 2026, the industry had completed a full user education cycle. Industry data shows that in 2025, the conversion rate of paid users for AI tools in China increased from 8% in 2024 to 11%. Among them, the willingness to pay among high-frequency users, such as professionals and creators, exceeded 30%, indicating that users have generally accepted the business logic of "free basic functions and paid high-value productivity capabilities." Entering the market at this point, ByteDance no longer needs to do the thankless work of market education; it only needs to convert users from its existing pool of paying users. The difficulty of the domestic C-end subscription business is obvious: a 30% annual renewal rate for tool products is already top-tier in the industry; price wars are the norm, and the cost of switching users is almost zero. The core challenge in the large-scale paid business lies in balancing "paid revenue" with "computing costs," and ByteDance possesses a leading technological advantage in this area domestically. The essence of large-scale commercialization is calculating whether revenue can cover computing costs. This is an industry-wide challenge. Users willing to pay are often the most heavy users; and high-frequency use means higher computing power consumption. ByteDance's advantage lies in its leading position in China's model efficiency and cost control. According to publicly available technical data, Doubao 2.0 achieves a 43% improvement in inference efficiency, reduces first-packet latency in long-context scenarios by more than 25% compared to mainstream industry models, and boasts a 99.98% success rate in high-concurrency scenarios, placing it among the industry's top performers in stability. Simultaneously, its 10,000-token inference cost is only 38% of that of leading overseas models operating within a compliant domestic network. This significant cost advantage allows it to better support the stable operation of paid, high-computing-power tasks. However, several structural problems in the domestic C-end subscription market remain unresolved by any player, and ByteDance is no exception. The first problem is that users are willing to pay, but not willing to keep paying. The free-for-all mentality of the Chinese internet over the past two decades is ingrained in users' minds, a core challenge that all subscription products must face. Even now, a 30% annual renewal rate for domestic C-end tool products is considered top-tier in the industry, while overseas similar products generally have renewal rates above 60%. The core reason is that domestic users mostly pay for "emergency" services: they buy a one-month membership temporarily to write a proposal or work on a project, and then immediately unsubscribe afterward, lacking a long-term habit of continuous payment. More importantly, ByteDance has no successful experience in this area. ByteDance's past commercial success has relied primarily on advertising, e-commerce, and live-streaming rewards, not on To C subscription businesses. Even the premium memberships for apps like CapCut and Douyin are only supplementary revenue streams; they have never created a tiered subscription system for a national-level core product. Faced with the industry challenge of "low willingness to pay continuously" among domestic users, ByteDance has no ready-made solution, which is the biggest unknown. The second problem lies in the "substitutability" of paid value. Currently, the core paid features disclosed by Doubao include long document reading, PPT generation, in-depth data analysis, and batch high-resolution image generation. These are essentially industry-standard capabilities—most of the free versions of domestic competitors already offer basic services for these features, and many open-source models can be implemented completely free of charge with local deployment. If Doubao's paid version can only provide "faster response speed, more calls, and slightly improved model capabilities," without delivering a crushing experience improvement or irreplaceable exclusive value, users will not develop a sustained willingness to pay. Even with ByteDance's ecosystem integration, if it's merely a superficial functional integration rather than a truly integrated, closed-loop experience embedded in the user's production workflow, it cannot create genuine monetization. Ultimately, it will likely suffer from the industry-wide problem of "enthusiastic initial subscriptions followed by high cancellation rates in the second month." The third problem lies in the "bottomless pit" of computing power costs. This is the most fundamental and intractable contradiction. Theoretically, more paying users mean higher revenue; however, in AI, more paying users can also lead to a simultaneous or even faster increase in costs. If usage isn't restricted, paid revenue might not even cover costs, creating a vicious cycle of "the more paying users, the greater the losses." Restricting usage, on the other hand, directly damages the paid user experience, leading to user dissatisfaction and a collapse in reputation, resulting in the dilemma of "restricting usage and losing users, not restricting usage and losing costs." This is an almost unsolvable balance problem. Even ChatGPT couldn't solve it—in 2024, ChatGPT's operating losses exceeded $5 billion, and subscription revenue was far from covering costs. Even if ByteDance has stronger cost control capabilities, it will be difficult to completely escape this industry-wide vicious cycle. Finally, a price war will inevitably return. The competitive logic in the domestic market is simple: as soon as a product proves it "can make money," other players will quickly follow suit and then undercut prices. Once competitors launch a new round of price cuts, subsidies, and free memberships—a fiercely competitive offensive—Doubao will face a dilemma: follow suit and fall into the "low-price trap" it deliberately avoided, disrupting its original pricing system and cost model; not follow suit and face a large-scale loss of price-sensitive users. ByteDance has won many battles in the past by relying on low prices and subsidies, but this time, it's setting higher prices. Whether it can withstand the fierce competition and insist on value-based pricing is a huge test.
The beginning is not difficult, the long run is difficult
Returning to the core question: Can Doubao's paid subscription model succeed?
The answer is actually quite clear. Achieving a large number of paying users and initial commercialization in the short term is highly probable; however, whether it can establish a healthy profit loop and become a benchmark for C-end AI subscriptions in China in the long run is full of uncertainty.
In the short term, with a base of 345 million monthly active users, even if only a 1% paid conversion rate is achieved, it can quickly form a scale of 3.45 million paying users.
In the long run, however, whether this venture can truly succeed depends not on its user base size, its model capabilities, or even the extent of ByteDance's ecosystem advantages, but on whether it can solve two fundamental problems: First, can it truly transform ByteDance's ecosystem advantages into irreplaceable, mandatory user payments, rather than just a marketing gimmick? Second, can it break free from the vicious cycle of "low-price competition, low renewal rates, and cost inversion" in the domestic C-end subscription industry and find a truly healthy and sustainable business model? For the entire domestic AI industry, Doubao's commercialization attempt has a benchmark significance far exceeding the product itself. If Doubao succeeds, then the C-end commercialization of the domestic large-scale model industry has found a feasible path that does not rely on low-price competition; if it ultimately fails due to the industry's structural problems, then domestic large-scale model players need to rethink: how can the C-end subscription path truly be successful?
Catherine
Weatherly
Kikyo
Anais