Large model vendors such as MiniMax open stores on Tmall to sell Token packages

By Photon Planet

On the eve of Double 11, large model vendor MiniMax opened its first official brand flagship store on Tmall, selling Token subscription packages. The store has now listed the M Plan subscription packages, including three tiers—Go, Explore, and Build—available for monthly or annual purchase.

(Source: Tmall platform)

This model of listing Token subscription packages on e-commerce platforms is similar to the logic of e-commerce selling mobile data packages and phone credit top-ups in earlier years: carriers quantified and packaged invisible, intangible resources into prepaid products that users select monthly and that are deducted as used. The only difference is that the transaction object has changed from communications data to large model inference compute.

In addition to MiniMax, in September this year, large model vendors such as Zhipu, Alibaba Cloud Model Studio, and Qianwen Office have all opened flagship stores, and Tmall launched the "AI Space Station" top-up center in the same period, placing the subscriptions of large model vendors under the same entry point.

Behind AI vendors' bustling entry into e-commerce platforms is the shift of large model API call quotas from backend technical interfaces to more standardized digital goods.

Over the past two years, a core metric in large model competition has been technical capability, and vendors have focused more on and maintained B-end customers and professional developers for private deployment business. Now, beyond specialized customer channels, sales channels for AI model invocation services are gradually opening up.

At present, Tmall has officially confirmed that it has communicated with large model companies, mainly about what kind of bundle sales there will be for Double 11, according to different demand types—study type, family type, one-person studio type, and so on—hoping to cover as many distinctive groups as possible.

Selling Tokens Is Replicating the Data Package Model

A Token is the unit of measurement for information processed by large models and also the billing unit for AI services.

The model reads requests, performs inference, and generates answers, and every step consumes Tokens, just like a mobile phone data package: every message sent and every video watched consumes data. When large model vendors sell packages on e-commerce platforms, they are essentially packaging and selling Token invocation quotas.

On September 2, Zhipu officially opened an official flagship store on Tmall, putting subscription packages such as Coding Plan on sale. On September 3, Tmall launched the "AI Space Station (Token Top-up Center)," with the first batch connecting domestic large model vendors such as Alibaba Cloud, Zhipu, Kimi, MiniMax, and DeepSeek.

As of now, Zhipu, Alibaba Cloud, and MiniMax have all opened official flagship stores, while Kimi and DeepSeek are represented by the Luquan flagship store.

The two sales models reflect a vendor's pace of development.

Vendors using the agency model only need to authorize distribution and do not need to handle pricing, delivery, peak-period operations and maintenance, and other issues. Opening a self-operated flagship store, by contrast, means the vendor must independently handle product listing, customer service, pricing strategy, and marketing activities, while also completing platform qualification review, which places new demands on the vendor's brand communication capabilities. Some media reports said that over the past few months, multiple model vendors have intensively recruited brand communication teams, and Zhipu recently also began hiring a "brand narrative manager."

At the product level, the product packages launched by major vendors share many common features.

In terms of pricing, vendors with official flagship stores have all set three or more price tiers and offer personal and team editions, supporting monthly, quarterly, and annual subscriptions.

In terms of product rules, such subscription packages generally retain a dual rate-limiting mechanism consisting of a 5-hour rolling window and a weekly quota to ease peak compute pressure.

Essentially, by moving Token subscription packages from official websites to e-commerce platforms, large model vendors are trying to replicate the data package model of the internet era and the entire rich media ecosystem derived from data as a foundation.

In the early days of the mobile internet, mobile data was billed by volume in KB. At that time, data fees were high and billing had no cap, and users loading rich media content such as images and videos could easily incur high charges, which constrained data consumption.

As communications infrastructure was completed, the marginal cost of transmitting data fell sharply, and carriers successively launched different data package plans, under which users could obtain data quotas by subscribing. The data package model offered individual packages, family packages, as well as directional data, idle-time data, and general data, thereby meeting differentiated data needs across scenarios and achieving efficient allocation of data resources.

This package model locked in the upper limit of data consumption, spread out the per-unit data cost, and spurred massive demand for graphic and short-video content, from which mobile internet application ecosystems such as WeChat and Douyin also emerged.

A similar transformation is now unfolding in the large model compute market.

When major carriers moved the sale of data packages from offline business halls to e-commerce platforms, the core logic behind their consideration was to capture new user numbers and network subscriptions. Today, when large model vendors move Token packages from exclusive official website channels to e-commerce platforms with large numbers of C-end consumers, their core purpose is also to establish cognitive reach and brand perception with C-end users.

For large model vendors, e-commerce platforms are not only a channel for facilitating transactions but also a marketing window aimed at C-end consumers.

As an e-commerce platform, Tmall has hundreds of millions of C-end consumers. In the past, model companies' user groups pointed more to B-end companies, G-end governments, and D-end developers. When they enter e-commerce platforms, they can reach C-end users more precisely, and every shopping search and page browse by consumers is an opportunity for model vendors to occupy consumers' minds.

In terms of package segmentation, the product segmentation of Token packages is quite similar to that of data packages. Token packages target "study type, family type, one-person studio type" and so on.

Taking MiniMax's M Plan subscription package as an example, it is divided into three tiers: Go, Explore, and Build, and supports monthly and annual purchases. This kind of product segmentation can meet the different needs of C-end users from light to high-frequency use.

When model computing power becomes a standardized, easily purchasable, and independently subscribable digital commodity, large model computing power will surround our lives like electricity and water. What awaits next is more people using it and exploring applications and ecosystems belonging to the Token world.

Large models move down to e-commerce

As large models accelerate their deployment across thousands of industries, domestic Token call volume is rising rapidly. As of June 2026, China's average daily Token call volume exceeded 500 trillion, a more than thousandfold surge in two years compared with 100 billion at the beginning of 2024.

Huge incremental demand is spawning fierce market competition.

In March this year, Alibaba announced the establishment of the Alibaba Token Hub (ATH) business group, with "creating Tokens, delivering Tokens, and applying Tokens" as its core goals, directly led by group CEO Wu Yongming. In the same month, Tencent Cloud upgraded its MaaS platform to Token Hub. In May this year, the three major telecom giants also collectively entered the market, launching Token packages covering both individual and government-enterprise users.

From cloud vendors to telecom giants and then to independent large model vendors such as Zhipu and MiniMax, Token has become the hottest business in the AI industry chain.

Zhipu's financial report for the first half of 2026 shows that Token call-related revenue was 825 million yuan, a year-on-year increase of 2735.7%, and its share of total revenue soared from 15.2% in the same period last year to 86.5%. The Token call business has become Zhipu's revenue engine.

At a time when most large model vendors are still loss-making and investment far exceeds returns, the capital market increasingly values real deployment revenue.

Vendors developing the Token business can directly generate cash flow and prove their commercialization potential to the capital market. At the same time, large model vendors also hope to use e-commerce platforms to launch more diverse Token package usage scenarios and create a bigger commercial pie.

However, judging from the current situation, the participation of large model vendors in this Double 11 event is still more of a marketing gimmick than actual sales, and whether greater potential can be unleashed later remains a question mark.

On e-commerce platforms, to meet C-end demand, large model computing power services are evolving toward standardized, mass-market product forms.

The first shift is a lower barrier to use.

In the past, Token purchases were aimed at professional developers and required users to understand technical terms such as API Key and console.

Taking MiniMax as an example, users need to log in to the MiniMax open platform, complete real-name verification, register a developer account, properly keep the API Key and Group ID, and then fill the key into code, local CLI tools, or third-party AI clients. Only after completing this entire process can they call the model.

This usage model splits "buying Tokens" and "using Tokens" into two separate things. For non-technical users, every step in the process filters out a group of people.

But after vendors enter Tmall, users only need to select a package in the Tmall store, complete payment, and log in with one click to bind their personal account.

In terms of purchase channels, users no longer need to navigate one platform after another, but can buy directly on everyday e-commerce. In terms of professional knowledge, technical concepts have been replaced by general terms such as monthly cards, annual cards, and continuous subscriptions. Overall, from targeting professional developers to entering e-commerce platforms to target C-end consumers, Token package products are continuously developing toward a lower barrier to entry.

Although the barriers to purchase and use have been lowered, sales data from the official flagship stores of major vendors are not so satisfactory. Zhipu's official flagship store lists a total of 6 products; the best-selling product sold more than 200 units, while the lowest sold only one.

(Source: Tmall platform)

The second shift is that Token packages are becoming more standardized.

After large model vendors list their Token packages on e-commerce channels such as Tmall and JD.com, they need to follow the platforms' mature commodity trading rules. In such an e-commerce sales scenario, Token packages are transformed into standardized digital commodities with transparent terms and clear rights and responsibilities.

Before sale, e-commerce platforms need to first review vendor qualifications to ensure that the delivery chain is real and controllable. After listing, vendors also need to clearly disclose product-related information as required.

For example, Alibaba Cloud's Token Plan package will mark complete product introductions, supported models, supported Agent tools, package usage, redemption process, precautions, and other information on the product details page for users to check. During the sales process, customer service is available at any time to answer consumers' questions, and product prices can also participate in platform promotional activities.

Compared with after-sales service for package products on official websites, Token packages listed on e-commerce platforms follow the e-commerce after-sales system and come with rights such as order retention and after-sales consultation. In addition, the "user reviews" on e-commerce platforms will also provide timely feedback on more authentic package usage experiences.

At present, user experience records, good or bad, have already appeared in the official flagship stores of various vendors, which may affect the public's brand perception of the vendors.

In addition, as a new e-commerce category, the measurement standards for Token packages listed by various vendors are not unified.

On the Tmall AI Space Station, for packages at the same lowest tier, Qwen's Lite version is 39 yuan for 11,500 Credits/month, Zhipu's Lite version is 118 for 10,000 points per week, and MiniMax's Go version is 49 yuan, with the quota only marked as "more usage" and no specific quota indicated. For C-end users, this still feels like a technical black box.

Token packages remain a niche product within a professional circle. Although vendors have launched products for e-commerce C-end channels, C-end sales scale has not yet expanded, and Double 11 store sales are also below expectations.

Whether large model vendors moving down to e-commerce is a C-end trial or a sustained commercial structural transformation remains to be seen.