AI News

Major model companies such as Zhipu, MiniMax, Kimi, and DeepSeek clustered onto Double 11 to list Tokens, most without real price cuts

36Kr · AIUpdated Nov 16 19:16

AI summary

Major model companies such as Zhipu, MiniMax, Qwen Office, Alibaba Cloud, Kimi, and DeepSeek clustered onto e-commerce platforms during Double 11 to list Token packages, but most stores offered no discounts; Zhipu's GLM Coding Plan Lite monthly card dropped only from 118 yuan to 115 yuan. On the first day of Tmall Double 11, Token category transaction value grew 450% month-over-month, but cumulative sales at Zhipu and MiniMax flagship stores remained only in the hundreds. As of June 2026, China's daily average Token calls had exceeded 500 trillion, and vendors increasingly treat e-commerce as a comparison and marketing window.

Body · AI translation

This Double 11, large model companies such as Zhipu, MiniMax, Qianwen Office, Alibaba Cloud, Kimi, and DeepSeek are all crowding into e-commerce to sell Tokens.

Many consumers say, "Buy Tokens at 50% off on Double 11" and "AI New Year goods can be stockpiled."

But Crazy Tech Hall asked customer service at multiple stores including JD Cloud, Tmall MiniMax, and the Alibaba Cloud flagship store, and most stores currently have no Token products participating in Double 11 discounts; Zhipu is one of the few actually offering a discount, with the GLM Coding Plan Lite monthly card dropping from 118 yuan to 115 yuan, a savings of only 3 yuan.

(Zhipu JD flagship store)

(Zhipu official website)

Since they do not plan to rely on price cuts to grab sales, why are large model companies still rushing to open stores collectively on Double 11?

01> Is buying Tokens worth it? First, let's see what consumers are actually getting

Open Taobao and search for Token, and you can directly see the Tmall AI Space Station, which integrates products from vendors such as Qianwen, Zhipu, and MiniMax. Monthly cards, quarterly cards, annual cards, Lite, Pro, Max, with prices ranging from tens of yuan to over a thousand yuan, and how much the discount is is also clear at a glance.

At first glance, buying Tokens is similar to topping up a phone bill or buying game point cards.

Taking Zhipu as an example, several GLM Coding Plan products are sold at almost the original price.

(Zhipu Taobao flagship store)

MiniMax has also listed the M Plan package on Tmall, with some monthly cards at 50% off for the first order during the opening period, starting at 24.5 yuan.

(MiniMax Taobao flagship store)

Beyond price, we also looked at user feedback after purchase.

In the Zhipu JD flagship store, the positive review rate shown on a monthly card product page is only 67%. Looking through the comments, some say the top-up process is not fully connected and requires manual assistance; others complain that top-up is too slow, directly pointing to a poor experience.

(Post-purchase comments under the Zhipu JD store)

In addition, there is another kind of third-party relay service.

For example, we randomly selected a store that can help you top up and access various major models connected to DeepSeek. You can choose to directly buy the v4.1Flash monthly package, which clearly labels 100 million or 1 billion tokens.

But there is another type that lets you use all models, though its purchase unit is "quota."

(Reply from customer service at a third-party top-up store on JD)

After communicating with customer service, we found that this "quota" is hard to convert into tokens, and we also do not know how many tasks I can complete after buying this package.

Moreover, different relay providers do not use a unified billing method. Some use custom quotas or points, and some charge by number of requests, but none can clearly explain how many Tokens these numbers correspond to.

02> Why has Token consumption suddenly become hot?

As of June 2026, China's average daily Token calls had exceeded 500 trillion. By comparison, at the beginning of 2024 this number was about 100 billion, and by the end of 2025 it reached 100 trillion. That is, in more than two years, average daily calls increased by about 5,000 times; in the first half of 2026 alone, they grew from 100 trillion to 500 trillion.

The "Generative Artificial Intelligence Application Development Report (2026)" released by the China Internet Network Information Center also shows that as of the first half of 2026, the number of generative artificial intelligence users in China had exceeded 700 million, with a penetration rate of over 50%. In addition to having AI answer questions, 47.8% of users use it to process images or videos, and 32.5% of users use it to handle work tasks such as work summaries, meeting minutes, and PPTs.

The user base has expanded, and usage scenarios are also increasing.

This also explains why the revenue of large model companies is increasingly coming from model invocation services.

Taking Zhipu as an example, according to performance data for the first half of 2026, the company's revenue in the first half reached 954 million yuan, a year-on-year increase of 399.7%. Among this, revenue from the open platform and API business reached 825 million yuan, a year-on-year increase of 2,735.7%, accounting for 86.5% of total revenue.

MiniMax has shown a similar change. According to the company's 2026 interim results, in the first half of the year revenue from the open platform and other AI-based enterprise services reached 73.9 million US dollars, a year-on-year increase of 703.1%, accounting for 63.4% of total revenue. The company also disclosed that as of July 2026, platform Token consumption had reached 20 times that of January of that year.

When AI changes from an occasionally used chat tool into a work tool that requires continuous invocation, Tokens have the foundation to be sold separately.

Monthly cards and package bundles are exactly a way for vendors to try to turn model invocation capabilities into standardized goods.

03> No price cuts, so why still join the Double Eleven excitement?

But if Tokens really were a mature e-commerce category, vendors should try to cut prices, create bundles, and push sales. Yet what we saw this Double 11 is that brands are flocking in, but the promotional intensity is not great.

According to an October 23 report by Yicai, on the first day of Tmall Double 11, transaction value in the Token category increased by 450% month-on-month.

This number looks very intimidating.

We opened JD, and the Zhipu JD self-operated flagship store has only 355 followers, while the JD Zhipu GLM Coding Plan Lite monthly card shows only over 100 units sold.

The Tmall Zhipu flagship store has cumulative sales of about 300 units across all Token products.

The MiniMax flagship store has also sold around 200-plus units since launch.

Looking at it this way, this 450% growth is more like a new category beginning to be seen by consumers, rather than it already having become a hit that everyone is buying.

On one side, product sales are still at the hundred-unit level; on the other side, mainstream vendors are accelerating their entry and platforms are opening dedicated service entrances.

Compared with how many monthly cards they sell right now, they are clearly still striving for something else.

Tmall President Jialuo said this in an interview at the Double 11 launch event: let consumers be able to compare different large model solutions in one stop on Tmall. Tmall is also communicating with vendors to design Token combination packages according to needs such as learning, family, and one-person studio. Jialuo also mentioned that after large models enter the platform, the attention they receive will increase significantly, and Tmall is not just a trading platform but also like a marketing window.

For vendors, in the past, ordinary users usually came into contact with large models by opening some AI application; developers buying model capabilities might have to go to the official website to register an account, read API documentation, and configure invocation methods. Now, model subscriptions and Token packages have been moved into Taobao and JD, and users can place orders just like buying other digital services.

But after all, Tokens are not like phones or earphones, where users can intuitively see hardware specifications; their value must be demonstrated through actual invocation. Whether package rules are clear, whether top-up is convenient, and whether price matches consumption will all affect whether users are willing to continue buying.

Otherwise, Tokens may first become a very lively e-commerce category, yet long fail to become a business that makes consumers willing to keep paying.

This article comes from the WeChat public account "Crazy Tech Hall", author: Kong Yilin, published with authorization from 36Kr.

Source

36Kr · AI